Reproductive Biology and Relative Male and Female Fitness in a Trioecious Cactus, Pachycereus pringlei (Cactaceae)

by Theodore H. Fleming, Sandrine Maurice, Stephen L. Buchmann, Merlin D. Tuttle
Reproductive Biology and Relative Male and Female Fitness in a Trioecious Cactus, Pachycereus pringlei (Cactaceae)
Theodore H. Fleming, Sandrine Maurice, Stephen L. Buchmann, Merlin D. Tuttle
American Journal of Botany
Start Page: 
End Page: 
Select license: 
Select License

American Journal of Botany 81(7): 858-867. 1994.




Department of Biology, University of Miami, Coral Gables, Florida 33124;
Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721;
USDA, ARS, Carl Hayden Bee Research Center, 2000 E. Allen Road, Tucson, Arizona 85719; and
Bat Conservation International, P.O. Box 162603, Austin, Texas 78716

We describe the breeding system of an autotetraploid trioecious cactus, Pachycereus pringlei, provide estimates of the fitnesses of males and females relative to that of hermaphrodites, and discuss the role played by pollinators in the maintenance of three sexual morphs. Relatively high frequencies of females (45%) and males (26%) exist in coastal desert populations around Bahia Kino, Sonora, Mexico. They differ from hermaphrodites in flower size (females only), initiation of the flowering season, number of flowers produced per night and per season, sucrose content of nectar, and, in females, number of fruits produced per season under open pollination and in response to hand-pollination. Major similarities between the sex classes include overall plant size, nectar volume per flower, percent fruit set in open-pollinated flowers offemales and hermaphrodites, seed mass and number of seeds per fruit, and pollen mass per flower in males and hermaphrodites. Hermaphrodites are self-compatible, and the selfing rate is high (65%). Levels of inbreeding depression in selfed fruits and seeds appear to be low. Fruit set is strongly pollinator-dependent in females but much less so in hermaphrodites. Relative fitness of males and females, as measured by annual production of pollen or seeds, is at least 1.5 times higher than that of the corresponding sex function in hermaphrodites. Given the high selfing rate and apparent lack of inbreeding depression, these fitness differences are insufficient to explain the occurrence of trioecy in this species.

Most angiosperms (ca. 72%; Yampolsky and Yampol- sexual types apparently resulting from the breakdown of sky, 1922) are hermaphrodites, i.e., have the ability to dioecy and reversion to hermaphroditism. Despite its rar- transmit genes through both male (pollen) and female ity in angiosperms, trioecy is of considerable theoretical (seed) function. Relatively few angiosperms are dioecious interest because, as pointed out by Gregorius, Ross, and or contain mixtures of unisexual and bisexual individuals Gillet (1983), it can shed considerable light on the evo- (i.e., gynodioecious, androdioecious, or trioecious). Trioe- lution of other sexual systems in plants, particularly dioe- cy-the co-occurrence of males with seed-producing fe- cy. males and hermaphrodites-is an especially uncommon Models of the evolution of nonhermaphroditic systems sexual system. Putative examples include a subspecies of in plants indicate that four parameters-pollen and seed Silene acaulis in the French Alps that has been reported fertilities of males and females relative to hermaphrodites, to contain three morphological sex-types (Bock, 1976). levels of self-fertilization, and degree of inbreeding de- Observations of fruit production in hermaphrodite in- pression-should play critical roles in determining wheth- dividuals, however, indicate that its reproductive system er males and females can invade populations of her- is better described as subdioecious, i.e., hermaphrodites maphrodites (e.g., Lewis, 194 1; Lloyd, 1975; Charnov, are really "inconstant" males (S. Maurice, unpublished Maynard Smith, and Bull, 1976; Charlesworth and data). The most carefully studied plant species with more Charlesworth, 1978). Phenotypic models and models based than two sexual morphs is Thymelea hirsuta (Dommte, on the nuclear determination of sex in outbreeding pop- Bompar, and Denelle, 1990). This species does not possess ulations indicate that the fertility of a unisexual must be a clearly trioecious reproductive system, but rather what at least twice that of the corresponding sex function in the authors called tetramorphism in which there are four hermaphrodites for it to become established in a her-

maphroditic population. Selfing in hermaphrodites re- Manuscript received 27 September 1993; revision accepted 7 Feb- duces the relative fertility needed by females, particularly ruary 1994. when there is a cost to inbreeding, but increases the rel-

The authors thank C. and P. Cabrera and J. Serracho for permission

ative fertility needed by males to invade populations of

to study plants on their property; P. Boyles and W. Hutchinson for

hermaphrodites. The presence of one unisexual type in a

hospitality in Bahia Kino; P. Benetiz and M. Rosenzweig for the loan

population makes it easier for the second unisexual type

of field vehicles; M. Horner, T. May, C. Sahley, V. Sosa, M. Schanz, B. Watkins, and others for assistance in the field; personnel at Bat Con- to establish (Charlesworth and Charlesworth, 1978). Pre- servation International, especially P. Morton and J. Debelak, for logistic dictions from models with cytoplasmic or nuclear-cyto- help; J. Bronstein, J. Kohn, and L. McDade for constructive reviews of plasmic sex determination are quite different because any previous drafts of this paper; and D. Charlesworth for her interest in

increase in the relative fertility of females compared with

and suggestions regarding this study. Financial support was provided

hermaphrodites can lead to their maintenance (Lewis,

by the National Geographic Society, National Fish and Wildlife Foun-

194 1 ;Charlesworth, 198 1 ; Delannay, Gouyon, and Val-

dation, and the National Science Foundation. S. Maurice was supported

by a grant from the French government. deyron, 198 1; Gouyon, Vichot, and Van Damme, 199 1). Author for correspondence (FAX: 305-284-3039). Nuclear and phenotypic models conclude that the co- existence of three sexual phenotypes is not expected, ex- cept under certain genetic constraints. The model of evo- lution of dioecy proposed by Charlesworth and Charlesworth (1978) indicates that hermaphrodites and doubly sterile individuals will be retained in "dioecious" populations if the male-sterility and female-sterility loci are not tightly linked, although such a linkage will be selected. Gregorius, Ross, and Gillet (1983) studied a one locus-two allele model of sex determination and showed that trioecy can be stable only when hermaphrodites are the heterozygous genotype. Recent models investigating the possibility of establishment of males in gynodioecious populations with nuclear-cytoplasmic sex determination show that trioecy can result from nuclear-cytoplasmic interactions (Gregorius and Ross, 1987; Maurice et al., 1993). Trioecy could also result from the breakdown of dioecy with hermaphrodites invading, but not fully re- placing, populations of males and females.

In this paper we describe the breeding system of a tri- oecious cactus, Pachycereuspringlei, provide estimates of the fitnesses of males and females relative to fitnesses of hermaphrodites, and discuss the role played by pollinators in the maintenance of three sexual morphs. The discovery of trioecy in the Cactaceae is unexpected because non- hermaphroditic sexual systems are apparently uncommon in this family of about 1,600 species. Only three species of Opuntia (0. stenopetala, 0. grandis, and 0. glauces- cens) and Echinocereus coccineus are known to be dioe- cious; Selenicereus innesii may be gynodioecious; Mammillaria dioica and M. neopalmeri may be gynodioecious or trioecious; and bisexual, dioecious, and trioecious pop- ulations occur in Opuntia robusta (Parfitt, 1985; del Cas- tillo, 1986; Hoffmann, 1992). Prior to our study, P. prin- glei was considered to be a hermaphrodite (e.g., Shreve and Wiggins, 1964) because all reproductive individuals appear to have perfect flowers. Our study, however, in- dicates that populations in central Sonora, Mexico, con- tain four sex classes-hermaphrodites, male-steriles (fe- males), female-steriles (males), and a few male and female steriles (neuters). In contrast to many subdioecious plants (Lloyd and Bawa, 1984), hermaphrodites in P. pringlei are not rare, and they set substantial fruit, i.e., they are not "inconstant males." Hence, this species represents a true case of trioecy. Calculations based on our estimates ofthe relative fitnesses ofthe three sex types and estimates of rates of self-fertilization do not support the hypothesis that this reproductive system represents a transitional stage toward dioecy.


Theplant-Pachycereuspringlei (Engelm.) Britt. & Rose, "cardon," is a large (its height can exceed 15 m), auto- tetraploid columnar cactus restricted to the desert por- tions of the Mexican states of Sonora and Baja California Norte and Sur (Murawski et al., in press). Its flowering season lasts from late March through early June with a peak in late April to mid-May. Its large white flowers open shortly after sunset and close by noon the next day. At night they are visited by the nectarivorous bat Leptonycteris curasoae (Phyllostomidae: Glossophaginae) and small moths. Beginning around sunrise, flowers are visited by several species of birds and bees. Pollinator exclusion experiments indicate that both nocturnal and diurnal vis- itors are important pollinators, although the nectar- and pollen-rich flowers suggest coevolution with bats (T. H. Fleming, M. D. Tuttle, and M. A. Horner, unpublished data). Reproduction is entirely by seed, with Leptonycteris bats and several species of terrestrial mammals and birds serving as dispersal agents (V. Sosa, Xalapa, Mexico, per- sonal communication).

The study area -Fieldwork was conducted in the spring (April-June) of 1989-1 99 1 near the town of Nuevo Bahia Kino, Sonora, Mexico on the Gulf of California (latitude 28.5"N). This site is located in the Central Gulf Coast region of the Sonoran Desert and is dominated by Bursera microphylla and two species of Jatropha (Shreve and Wig- gins, 1964). Other common plants include Larrea triden- tata, Cercidium microphyllum, Olneya tesota, and Fouquieria splendens. In addition to P. pringlei, three other species of columnar cacti- Carnegia gigantea, Stenoce- reus thurberi, and Lophocereus schottii-are common in the area; C. gigantea and S. thurberi are also pollinated by bats, birds, and bees whereas L. schotti is moth-pol- linated. Annual rainfall at Bahia Kino averages about 200 mm with most rain falling in July and August. Topography in the area includes gravelly flatlands punctuated by rug- ged hills 200-450 m above sea level. Our main study area, located about 9 km northeast of Nuevo Bahia Kino, en- compassed about 30 ha in which we marked, mapped, and sexed 143 individuals. Based on censuses of three

1-ha plots, adult P. pringlei density here was 7.7 plants per ha (range: four to 12 plants per ha). We also marked and sexed an additional 77 individuals in a dense "cardon forest" 15 km southeast of our main study area along Sonora state highway 16.

Sex ratios andflower and fruitproduction -To sex each marked plant, we dissected one or more open flowers and examined the anthers for pollen and the ovary for ovules. The presence of pollen could be scored unambiguously early in the evening, shortly after the flowers opened and the anthers dehisced. It was more difficult to score after sunrise because bats sometimes delivered large amounts of pollen to female flowers, causing a few to be initially scored as hermaphrodites. Ovule presence was less dif- ficult to score; ovaries of nearly all individuals contained either many tiny ovules or none. Repeated observations of many plants indicated that the sex of each plant was constant within and between years.

To document differences in flower characteristics of the four sex classes, we weighed and measured up to four open flowers per plant. Flower mass was determined to

0.1 g with a Pesola scale. The flower was then cut in half longitudinally and the following measurements were taken to the nearest mm with a plastic ruler: greatest overall length, internal width across the base of the tepals, length of the stigma and style, and internal length and width of the ovary. Summary statistics are based on only one flower per plant. Correlation analysis indicated that flower size, as indicated by mass and length, was concordant within pairs of flowers from each of 27 plants (mass: r2 = 0.83, P < 0.001; length: r2 = 0.68; P < 0.001).

Seasonal patterns of flowering and nightly patterns of nectar production were documented in 1989 and 1990 during the pollinator exclusion study (T. H. Fleming, M.

D. Tuttle, and M. A. Horner, unpublished data). Numbers of buds, open flowers, and developing fruit were counted weekly in 20 census plants (ten males, five females, four hermaphrodites, and one neuter); plants were chosen ran- domly before the species' sexual system was known. To more carefully document sex differences in the initiation of the flowering season, we scored 67 and 69 plants for the presence or absence of flowers on 6 April and 12 April 1990, respectively. On 22 May 1990 we measured the volume of nectar produced between 2 100 and 2300 hours (the time of peak nectar production) using a 3-cc syringe and determined the sucrose equivalent (weight/weight) of the nectar of one flower from 12 females, ten males, nine hermaphrodites, and two neuters using a temperature- compensated refractometer. To prevent pollinators from removing nectar, flowers were covered with bridal veil netting (mesh size = 1 mm) before opening.

The proportion of flowers setting fruit in females and hermaphrodites was determined for open-pollinated, hand-pollinated, and, in hermaphrodites, self-pollinated (unvisited) flowers in 1990. A series of potentially open- pollinated flowers was tagged during the pollinator ex- clusion experiments. The hand-pollination experiment involved thoroughly dusting the stigma of two tagged flowers with the anthers of one-fourth of a flower (bearing approximately 300,000 pollen grains) from a single pollen source in ten females and ten hermaphrodites. Flowers were covered with netting before and after hand-polli- nation. This experiment was repeated twice using the same group of females and hermaphrodites. Five flowers each from one male and one hermaphrodite were used in the first experiment; two males were the pollen donors in the second experiment. The self-pollination experiment in- volved covering two tagged flowers with netting shortly before they opened on a group of 13 hermaphrodites on two nights; netting was removed the next day after the flowers had closed. All flowers were tagged by inserting a piece of wire bearing a labeled aluminum tag into the stem surface 1-2 cm away from the flower. The fate of each flower (developing fruit present or absent) was scored at least 3 weeks after it was tagged. Most fruits that abort do so within 2 weeks after flowers close in P. pringlei.

To determine whether pollen source (male or hermaphrodite) influenced seed set in females and hermaphrodites, we conducted another hand-pollination ex- periment in 199 1. Twenty flowers on each of ten females and seven hermaphrodites were hand-pollinated as de- scribed above. Ten flowers were pollinated with one-fourth of the stamens from one of ten different males, and ten flowers were pollinated with one-fourth of the stamens from one of ten different hermaphrodites. Seven of the ten hermaphrodites served both as pollen recipients and pollen donors; three hermaphrodites served only as pollen donors. One flower per plant was pollinated with male pollen and one with hermaphrodite pollen on each of ten nights. Treated flowers were tagged, and fruits were scored as present or absent 3 weeks after the end of the experi- ment.

Seasonal fruit production in females and hermaphro- dites was determined by counting the number of maturing fruits on a series of plants in the last week of May in 1 990 and 199 1. To control for plant size, we also counted the number of branches 2 1 m in length on each plant in 1990. To determine the number of seeds per fruit, we collected two mature or nearly mature open-pollinated fruits from 15 females and 13 hermaphrodites in late June 1990. Seeds were removed from the fruits and air-dried before being placed in labeled envelopes. In the lab, un- filled seeds were separated from filled seeds and counted. The mass of the entire lot of filled seeds was determined to the nearest 0.1 mg, and two groups of 100 seeds were weighed to determine the average mass of a seed. Total seed mass divided by average mass per seed provided an estimate of the number of filled seeds per fruit. The ac- curacy of this estimate was determined by comparing the total and estimated seed counts for four fruits. The dif- ference between these counts averaged 0.6% (range: 0.4% to 0.9%).

To determine whether seeds from open-pollinated fruits produced by females and hermaphrodites in 1990 differed in their germination probabilities, we placed 50 seeds from one fruit each from 13 females and 13 hermaph- rodites on moistened filter paper in covered petri dishes. Dishes were kept at room temperature (ca. 25 C) under fluorescent illumination from 0730 to 1830 hours, and the number of germinated seeds was counted daily for 14 days. Seeds were scored as "germinated" when the radicle extended beyond the seed coat.

Pollen production -We determined the amount of pol- len produced by male and hermaphrodite flowers in 199 l by collecting one flower just before it opened from each plant used in the hand-pollination experiment. The bot- tom halfofthe flower containing the nectary was removed, and the top half containing the anthers was placed upside down on a labeled piece of aluminum foil overnight. The next morning, the flower was tapped vigorously to remove any pollen that had not fallen onto the foil, and the foil was folded and stored in a plastic bag in an ice chest until it could be processed in the lab. In the laboratory, the mass of pollen produced per flower was determined by weighing the foil to the nearest 0.1 mg before and after the pollen was removed. One weighed aliquot of 5 to 19 mg of pollen from each flower was acetolyzed and run through a HIAC-ROYCO model PS-320 particle size an- alyzer to determine the number of grains per mg of pollen as described by Buchmann and Shipman (1990). The number of pollen grains produced per flower was esti- mated by dividing total pollen mass by the number of grains per mg.

Pollinator observations-Observations on pollinator abundance and behavior were conducted during the pol- linator exclusion study. Observations pertinent to this study included counting the number of Leptonycteris curasoae bats emerging at sunset from the Sierra Kino cave located 7.5 km west of our main study area five times in 1989 and eight times in 1990 and counting the number of pollinator visits to groups of three to 32 open flowers at night and during the day. Our nocturnal counts and observations were made with Noctron IV night vision scopes equipped with a 75-mm lens and located at least 18 m away from the cave or flowering plants. In 1990 most of our flower observations took place in our main study area but also included 2 or 3 nights of observations

TABLE1. Sex ratios in two populations of Pachycereus pringlei in So- nora, Mexico. Proportions are in parentheses.

Number of hermaph-Slte, year Females Males rod~tes Neuters Total

Cardon forest, 1990 2 5 14 3 8 0 7 7
  (0.325) (0.182) (0.494) (0.0)  
Bahia Kino, 1990 50 24 2 8 4 106
  (0.472) (0.226) (0.264) (0.038)  
Bahia Kino, 199 1 6 1 42 3 5 5 143
  (0.427) (0.294) (0.245) (0.034)  

at plants located within 1 km of the Sierra Kino cave and at the cardon forest site about 22 km from that cave.

To increase our sample of nocturnally monitored flow- ers of P. pringlei in 1990, we placed a small, white paper "bat detector" measuring 1.75 cm2 with a 0.5 cm hole in its center in a group of flowers. Detectors were slipped over the stigma and style just after the flowers opened. Bats pushed the paper aside as they plunged their faces into flowers to obtain nectar (see photos in Tuttle, 199 1). Three flowers each on eight to ten plants were treated on 3 nights in April. Status ofthe bat detector (intact =flower not visited; pushed aside = flower visited) was scored at 2400 and 0300 hours on these nights.

Statistical analyses were conducted on a microcomputer using Abstat (Anderson-Bell Corp., Arvada, CO) or SYS- TAT (SYSTAT, Inc., Evanston, IL). Homogeneity of var- iances in one-way ANOVAs was determined by Bartlett's test or Box's small sample F approximation. Data were transformed to their natural logarithms when the assumption of homogeneity of variance was not met. Per- centages were arcsine-transformed before being analyzed. Summary statistics are reported as means f1 SE.


Population sex ratios-In both populations, males and females made up at least 50% of the adult population, and neuters were very uncommon. Sex ratios in the two populations differed significantly (x2 = 9.2, df = 2, P = 0.01; neuters were excluded from this analysis) (Table 1). Females (43%) were most common at Bahia Kino, and hermaphrodites (49%) were most common at the cardon forest site. At Bahia Kino, many more adult plants flow- ered in 1989 and 1990 than in 199 1, but the sex ratio of flowering plants did not differ between 1990 and 199 1 (x2 = 1.42, df = 2, P = 0.70). This suggests that individuals of the different sex classes do not differ in their failure to flower in some years.

Floral and vegetative characteristics -Univariate and multivariate ANOVAs indicated that the sexes differed in floral characteristics (Table 2A). Females and neuters generally had smaller flowers than those of males and hermaphrodites. Corolla width differed least among the sex classes; flower mass and length of the stigma and style differed most among the sex classes. Results of a MANO- VA based on four floral traits (mass, total length, style length, and ovary area) and three sex classes (females, males, and hermaphrodites) allow us to reject the null hypothesis that the flowers of these classes are similar in overall size (Wilk's Lambda = 0.77, F8,I2, = 2.13, P = 0.038).

Flowers of females and hermaphrodites did not differ conspicuously in number of anthers, but anthers of fe- males lacked pollen and were noticeably smaller and darker in color (yellow vs. creamy) than those of the hermaph- rodites. In males, stigmas and styles were fully developed; however, although their ovaries usually lacked ovules, they were large and contained masses of funicular tissue. Ovaries of a few (four of 56) males contained a small number of ovules that prevented immature fruits from aborting, as occurred in all other males and neuters. Ripe fruits from two of these individuals contained only im- mature seeds. Hence, males cannot reproduce by seed, and their phenotypic gender (i.e., proportion of repro- ductive allocation to female function; Lloyd and Bawa, 1984) is zero. Flowers of neuters contained femalelike anthers and a malelike ovary.

Overall size of the plants, as indicated by the number of branches r 1 m in length, did not differ among the sex classes (Table 2D); sex expression does not appear to be size-dependent in P. pringlei.

Flowering biology-In 1990, males and females began flowering earlier in the season than hermaphrodites and neuters. On 6 April, 46% (N = 35) and 44% (N = 16) of the females and males had open flowers compared with 7% (N = 14) and 0% (N = 2) of the hermaphrodites and neuters (x2 = 6.8 1, df = 2, P = 0.03; hermaphrodites and neuters were combined in this analysis). On 12 April, percentages were 66%, 77%, 47%, and OO/o, respectively (x2= 3.30, df = 2, P = 0.19). Throughout the 1990 flow- ering season, the cumulative flowering curves of her- maphrodites and neuters trailed behind those of females and males in the census plants (Fig. 1).

We used the ratio of immature fruits: flower buds in the middle of the flowering season (30 April 1989, 5 May 1990, and 29 April 199 1) to determine whether relative position in the flowering season (e.g., early flowering vs. late flowering) was concordant between years within our phenology census plants. Our reasoning here was that plants with a high fruit :bud ratio in midseason had begun flowering earlier than those with a low fruit: bud ratio. Spearman rank correlation analysis using 19 plants cen- sused in all 3 years indicated that this ratio was signifi- cantly correlated in the 3 years (rs 2 0.65, P < 0.01 in the three comparisons). Relative position in the flowering season was concordant among years. These ratios indicate that the sexual bias among early-flowering plants that we documented in 1990 is likely to occur in other years.

The number of open flowers per night differed signif- icantly among the sex classes in 1990 and 199 1 (Table 2B). Males opened significantly more flowers per night than females and hermaphrodites in 1990; males and females differed significantly in 199 1. Only three sex class- es (females, males, and hermaphrodites) were compared in the 1990 ANOVA whereas all four classes were com- pared in 199 1. Data summarized in Table 2 suggest that nightly flowering levels were lower in 199 1 than in 1990. However, because these two sets of data were gathered somewhat differently, a between-years comparison is un- justified. The 1990 data were taken throughout the flow- ering season, and repeated counts of the same individuals


A. Floral characteristics Mass (g, wet weight)

Total length (mm)

Corolla width (mm)

Style length (mm)

Area of ovary (mm')

B. Nightly individual flower production 1990

C. Nectar characteristics Percent sucrose equivalents

Volume before 2,300 hours (cm3)

D. Plant size Number of branches > 1 m long

P value, one-wayFemales Males Hermaphrodites Neuters ANOVA

a Means with the same lowercase letter do not differ at P = 0.05 in a Tukey HSD test. Neuters were excluded from ANOVAs when N was <4.

were averaged before summary statistics were calculated, whereas the 199 1 data represent single counts in a larger group of plants taken in early May.

Based on the data from the phenology census plants, the sex classes also differed in seasonal flower production in 1990 (Fig. 1). By integrating the area under each flow- ering curve, we estimate that males produced an average of about 788 flowers per season, whereas females and hermaphrodites produced about 630 and 554 flowers per season, respectively.

Sucrose content of nectar, but not nectar volume pro- duced before 2300 hours, differed among the sex classes (Table 2C). The sucrose content of nectar was high (ca. 33% by weight) and was significantly higher in females than in males and hermaphrodites. Although sample size was small, neuters also appeared to produce richer nectar than males and hermaphrodites. Hermaphrodites pro- duced slightly, but not significantly, more nectar before 2300 hours than did females and males (Table 2C). Ad- ditional data on nectar production in P.pringlei indicate that sucrose content and production rates are concordant among flowers within plants and within plants between years (T. H. Fleming, unpublished data).

Fruit set and seasonal fruit and seed production-In 1990, fruit set in open-pollinated flowers was similar in females and hermaphrodites and averaged 35% (Table 3A). The 1990 hand-pollination experiment revealed that fruit set is more likely to be pollinator-limited in females than in hermaphrodites. In females, fruit set increased significantly from 35% in open-pollinated flowers to 73% in hand-pollinated flowers (x2= 9.8, df = 1, P = 0.0017), whereas fruit set did not increase with hand-pollination in hermaphrodites (Table 3A). Hermaphrodites can self- fertilize, but fruit set from bagged but unmanipulated flowers was significantly lower than that of open-polli- nated flowers (8.7% vs. 35%; two-tailed Fisher's exact test, P = 0.013). Five of eight (62.5%) hermaphrodite flowers that were hand-pollinated using another flower from the same plant set fruit in the 1991 experiments. Electrophoretic analysis of the seeds of open-pollinated fruits from 14 hermaphrodites indicated that 65% of their seeds were self-fertilized (Murawski et al., in press). Thus, self-fertilization appears to be common in hermaphro- dites of P. pringlei but requires one or more pollinator visits to maximize fruit set.

Total fruit set was significantly higher in females than in hermaphrodites in 1990 and 199 1 (Table 3B). The ratio






ermaphs. (554)

0 0 2 4 6 8 10


Fig. 1. Flowering curves of male (N = lo), female (N = 5), and hermaphrodite (N = 4) phenology census plants of Pachycereuspringlei in 1990. Week one began on 7 April. Numbers in parentheses indicate mean seasonal flower production, determined by integrating the area under each curve.

TABLE3. Fruit and seed set data in females and hermaphrodites of

Pachycereus pringlei. Data in B are 2 i1 SE. Sample sizes are in

parentheses (number of flowers in A, number of individuals in B).

P value, Parameter Females Hermaphrodites I-test

A. Proportion fruit set Open-pollinated, 1990 0.35 (55) Hand-pollinated, 1990 0.73 (40) Hand-pollinated, 199 1 0.76 (2 15)

Self-pollinated, 1990 -
Self-pollinated, 199 1 -

B. Fruit and seed production No. fruits/plant, 1990 103.1 k 13.6

No. fruits/plant, 1991 137.3 i21.8
No. filled seeddfruit 1,632.5 k57.4
Proportion filled seeds 0.96
Seed mass (mg) 5.2 i0.2


of fruit set in females :hermaphrodites in the 2 years was

1.8 and 1.5. In both sex classes in 1990, number of fruits was significantly correlated with number of branches (i.e., plant size) (females: r = 0.62, P = 0.000; hermaphrodites: r = 0.57, P = 0.002). Results of a 2 x 2 ANOVA using log(X + 1)-transformed data from 18 females and 13 hermaphrodites whose fruit crops were counted in 1990 and 199 1 indicated that differences between years within sex classes was not significant (F,,,, = 0.87, P = 0.36), whereas differences between sex classes within years was significant (F,,,, = 6.64, P = 0.012). In hermaphrodites, total fruits produced in the 1990 season was positively correlated with mean number of flowers per night (r = 0.70, df = 10,P = 0.0 12), suggesting that male and female functions covary in hermaphrodites.

Number of seeds per fruit, proportion filled seeds, and seed mass did not differ between females and hermaph- rodites (Table 3B). Fruits contained an average of 1,640 filled seeds representing 97% of the total seed set. Seed mass averaged about 5 mg in both sexes. Number of filled seeds per fruit did not differ between selfed and open- pollinated fruits of three hermaphrodites. Number of seeds averaged 1,662.5 and 1,686.3, for selfed and open-pol- linated fruits, respectively.

Results of the 1990 seed germination trials were influ- enced more by date of flowering than by sex class. Seed germination in fruits produced by plants flowering before 6 April was 78.8% in females (N = five plants, 250 seeds) and 80% in the single hermaphrodite (N = 50 seeds). Germination in fruits produced by plants flowering after 12 April was only 39% in females (N = four plants, 200 seeds) and 26% in hermaphrodites (N = nine plants, 450 seeds); the latter values differ significantly (x2= 10.53, df = 1, P = 0.00 12). Differences in germination of seeds between early- and late-flowering plants and within late- flowering females and hermaphrodites probably reflect degree of fruit and seed maturation at the time the fruits were collected (the third week ofJune) rather than seasonal or sex class differences in the germinability of mature seeds.

TABLE4. Results of the 1991 hand-pollination experiment. Data are X i 1 SE and were arcsine-transformed for the analysis in B.

A. Percent fruit set Pollen source

Sex class (N) Male Hermaphrodite

Female (I I) 77.8 i5.5 74.5 i6.7 Hermaphrodite (7) 34.6 i5.6 51.3 i8.8

B. 2 x 2 ANOVA Source df MS F

Pollen source 1 0.0537 0.59 ns Sex 1 1.912 20.8 I*** Pollen x Sex 1 0.126 1.38 ns Error 32 0.0919

*** P < 0.001; ns = not significant.

Overall, the absence of differences in seed production per fruit, seed size, and seed germination between females and hermaphrodites suggests that ifinbreeding depression occurs in hermaphrodites, it is not expressed by any of these measures of plant performance.

Pollen production-Total pollen mass per flower and number of pollen grains per mg of pollen did not differ in males and hermaphrodites (P > 0.70 in t-tests). Males produced an average of 469.6 f3.4 (N = 17) mg of pollen per flower and 3,540.7 f309.8 (N = 10) pollen grains per mg. Hermaphrodites produced an average of 45 1.1 k 4.9 (N = 13) mg of pollen per flower and 3,438.3 i

457.3 (N = 10) pollen grains per mg. Total number of pollen grains per flower thus averaged about 1.6 x 10, in both sex classes. In hermaphrodites, this yields a pollen : ovule ratio of about 947: 1. Although sample sizes are small, our data suggest that pollen mass may be more closely correlated with flower size in males than in her- maphrodites. For this analysis, we correlated pollen mass measured in 199 1 with flower mass and length data taken from the same individuals in 1990. In the combined male and hermaphrodite data, pollen mass was significantly correlated with both flower mass and length (mass: r = 0.44, df = 19, P = 0.047; length: r = 0.44, df = 19, P = 0.044). Considered alone, correlations in males approached significance, whereas they did not in hermaph- rodites (males: pollen vs. flower mass-r = 0.56, df = 9, P = 0.07; pollen vs. flower length-r = 0.54, df = 9, P = 0.08; hermaphrodites: pollen vs. flower mass-r = 0.38, df = 8, P = 0.29; pollen vs. flower length-r = 0.38, df = 8, P = 0.28). If sex class differences in the pollen mass- flower size correlation hold up with larger sample sizes, then pollen production apparently is independent of flow- er size in hermaphrodites but not in males.

Eflects of pollen and ovule source on pollination suc- cess -Results of the 199 1 hand-pollination experiment indicate that there were differences among sex classes in fruit set and pollen effectiveness (Table 4). As in the 1990 experiment, fruit set was significantly higher in females than in hermaphrodites. Pollen source (i.e., males vs. hermaphrodites) did not affect fruit set significantly in either females or hermaphrodites, although there was a trend for pollen from hermaphrodites to yield higher fruit set in other hermaphrodites than pollen from males. Re- sults of this experiment suggest that, in the absence of pollen limitation, females have higher fitness through seed production than hermaphrodites and that, given equal access to ovules, males and hermaphrodites do not differ in fitness, although hermaphrodites may have somewhat higher fitness than males when pollinating other her- maphrodites.

Relative fitness of females and males-Data for esti- mating the fitnesses of females and males relative to that of hermaphrodites through female and male function are summarized in Table 5.Open-pollinated females produce an average of 1.62 times more fruits or seeds per season than hermaphrodites. The hand-pollination experiment, however, shows that females can produce more fruits than that in the absence of pollen limitation. Potential relative fitness of females in this case can be calculated as the ratio of production of fruits per season times proportion fruit set via hand pollination divided by proportion fruit set via open pollination for females and hermaphrodites. By this formula, potential relative fruit set in females is 3: 1 (i.e., [I15 x 0.75/0.35]/[70.9 x 0.39/0.35]). We must view this value with caution, however, because we esti- mated maximum proportion fruit set by pollinating only a small subset of the flowers produced per season by females and hermaphrodites and do not know, for ex- ample, whether females can actually set 75% of all of their flowers as fruit in the absence of pollen limitation. Data on fruit set in females of dioecious angiosperms presented in Sutherland (1986) suggest that 75% fruit set is not an unreasonable value.

Although males and hermaphrodites produce similar amounts of pollen per flower, males produce substantially more flowers per night and per season. Pollen production in males is thus 1.52 times greater per season than that of hermaphrodites (Table 5).

Nocturnal pollinator observations -Counts of Leptonycteris curasoae departures from the Sierra Kino cave in 1989 and 1990 indicated that the number of bats was high (about 2,500) in early April and declined to very low values (<100) in May and June before rising again in July

H. Fleming, unpublished data). Bat visitation rates to
pringlei flowers declined as a function of distance from this cave. Before 28 April when visits dropped to zero, visitation rates at flowers averaged 2.69 visits per flower per hour within 1 km of the cave (N = 15 flowers on 2 nights), 0.30 10.3 1 visits per flower per hour in our main study area 7.5 km from the cave (N = 101 flowers on 8 nights), and 0.11 visits per flower per hour at the cardon forest site 22 km from the cave (N = 14 flowers on 2 nights).

The proportion of flowers equipped with "bat detec- tors" that was visited by bats in 1990 was similar in females and in males and hermaphrodites combined. In both groups, 35% of the flowers were visited (N = 37 flowers on 12 female plants; N = 48 flowers on a total of 16 male or hermaphrodite plants). This visitation prob- ability is the same as the probability of fruit set (35%) in open-pollinated flowers of females and hermaphrodites in 1990. These results indicate that bats are equally likely to feed at pollenless (female) and pollen-rich (male and hermaphrodite) flowers.

TABLE5. Estimates of the relative fitnesses of females and males in

Pachycereus pringlei. Data in 1 represent average values based on

the combined 1990-1 99 1 means.

Relative fitness

  1. Female/hermaphrodite: seeds per plant per season 200,184.2/123,483.5 1.62
  2. Male/hermaphrodite: pollen grains per plant per season
1.31 x 109/0.86 x lo9 1.52


Results of this study indicate that Pachycereus pringlei has a trioecious breeding system in coastal Sonora, Mex- ico. Unisexual morphs exist at relatively high frequencies in populations, and they differ from hermaphrodites in numerous reproductive traits, including flower size, ini- tiation of the flowering season, number of flowers pro- duced per night and per season, sucrose content of nectar, and, in females, number of fruits produced per season and response to hand-pollination. Major similarities be- tween the sex morphs include overall plant size, nectar volume per flower, percent fruit set in open-pollinated flowers of females and hermaphrodites, seed mass and number of seeds per fruit, and pollen mass per flower in males and hermaphrodites. Hermaphrodites are self-com- patible and the selfing rate is high. Levels of inbreeding depression in selfed fruits and seeds appear to be low, but more data are needed on seed germination rates and off- spring fitness before we can reach firm conclusions about this. Fruit set is strongly pollinator-dependent in females but much less so in hermaphrodites. Finally, relative fit- ness of males and females, as measured by annual pro- duction of pollen or seeds, is at least 1.5 times higher than that of the corresponding functions in hermaphrodites.

Many of the differences and similarities that we have documented in the sex morphs of P. pringlei also occur in trioecious populations of the cactus Opuntia robusta as well as in dioecious and subdioecious plants in general (e.g., Darwin, 1877; Bock, 1976; Solomon, 1986; Lovett Doust, O'Brien, and Lovett Doust, 1987; Shykoff, 1988; Sakai and Weller, 199 1; Molau and Prentice, 1992). Opuntia robusta is diploid and has diurnal, bee-pollinated flowers. del Castillo (1 986) reported that sex morphs differ in flower size and size of stamens and ovaries, length of flowering season and number of flowers produced per day and per season, nectar volume per flower, pollen per flow- er, and number of seeds per fruit. Rate of selfing in her- maphrodites is high (60°/o), and fruit set in hermaphrodites and females is not pollinator-limited. As in P. pringlei, level of inbreeding depression appears to be very low. Annual pollen production is 7.07 times higher in males, and annual seed production is 1.57 times higher in females than in hermaphrodites. del Castillo (1986) concluded that trioecy in 0.robusta is restricted to disturbed habitats and is a transitional stage to dioecy, which is an evolu- tionary stable reproductive system according to calcula- tions based on the model of Charnov, Maynard Smith, and Bull (1976).

Pollen vs. resource limitation -In P. pringlei, pollinator availability appears to affect the fitness of males and fe- males more strongly than it does that of hermaphrodites. For example, our hand-pollination experiments indicate that fruit and seed production is more strongly pollen- limited in females than in hermaphrodites. In the absence of pollen limitation, females potentially can set 75% of their fruits compared with a maximum of 51% in her- maphrodites. In the absence of pollinator visits, her- maphrodites can set about 8% of their fruits. Males and females, of course, have no fitness in the absence of pol- linator visits. These observations suggest that the fitness of unisexuals should be correlated with the abundance of pollinators. If this is true, then males and females should be common wherever pollinators are common, and her- maphrodites should predominate in populations wher- ever pollinators are uncommon. Our data on differences in the sex ratios in the Bahia Kino and cardon forest populations are consistent with this prediction. The for- mer population is much closer to a Leptonycteris bat roost, and its flowers have higher bat visitation rates than those in the latter; females are the most common sex class at Bahia Kino whereas hermaphrodites are most common at the cardon forest site. Results of a survey of geographic variation in sex ratios throughout the range of P. pringlei also support this prediction (T. H. Fleming and S. Mau- rice, unpublished data).

Annual seed and pollen production appears to be lim- ited to a greater extent by resources in hermaphrodites than in females and males, as originally proposed by Dar- win (1 877) and discussed in detail by Charlesworth and Charlesworth (1987). Instead of resulting in lower pro- duction of pollen and ovules per individual flower com- pared with unisexuals, this limitation results in lower total flower and fruit production per season. The absence of reduced pollen and ovule production in hermaphroditic flowers also occurs in the subdioecious Schiedea globosa (Sakai and Weller, 199 1).

The evolution of trioecy-One way in which trioecy can evolve is by the establishment of a male-sterility mutation at one locus and a female-sterility mutation at another locus in a population of hermaphrodites; if the loci are unlinked, then unfertile or neuter individuals may also be produced (Ross and Weir, 1976; Charlesworth and Charlesworth, 1978). Based on the notation of Charles- worth and Charlesworth (1 978), females can become es- tablished in a hermaphroditic population if their relative fertility (1 + k) is > 2(1 -sd) where s is the selfing rate and d the inbreeding depression. Both s and d can vary from 0 to 1.0. Males can establish in a hermaphroditic population if their relative fertility (1 + K) is > 2(1 sd)/(l -s). The condition for the second unisexual to increase in frequency in the presence of the first unisexual and hermaphrodites is: K > (1 + s -2sd)/(k + 2sd -s).

Our data for P.pringlei indicate that the relative fertility of females is 1.62 (i.e., k = 0.62), of males is 1.52 (K = 0.52), the selfing rate is 0.65, and there seems to be little inbreeding depression (d = O?). It can easily be seen that these values do not explain the existence of trioecy in this species. Even with a relative fitness of 3.1 in females (i.e., no pollen limitation), males cannot invade a gynodioecious population (0.52 < 1.14). We have also cal- culated the minimum inbreeding depression needed to fit our data to this model. With our observed fertility and selfing values, a minimum inbreeding depression of 0.29 is needed to explain the presence of females. In the most favorable conditions of dominance between alleles and linkage between genes (Charlesworth and Charlesworth, 1978), males could invade a gynodioecious population if the inbreeding depression is >0.83. The other nuclear model of the evolution of trioecy (Gregorius, Ross, and Gillet, 1983) is inappropriate because it is based on an unrealistic model of sex determination and predicts that trioecious populations should always contain more than 50°/o hermaphrodites, which is not the case at Bahia Kino and in other populations in Sonora and Baja California

(T. H. Fleming and S. Maurice, unpublished data).

The above models deal with diploid organisms, and P. pringlei is an autotetraploid (Murawski et al., in press). In diploids, dominance relations at the sterility loci and the recombination rate between these loci can affect the conditions for the second unisexual to become established and can also affect sex ratios at equilibrium (Charlesworth and Charlesworth, 1978). Although tetraploidy is likely to affect these values, it is not expected to affect the con- ditions for the increase of the first sterility allele, and the conditions given above for the invasion of the second unisexual type are the minimum ones. Therefore, our conclusions about the ability of unisexuals to invade her- maphroditic populations of this cactus are valid even for a tetraploid species.

Tetraploidy is known to sometimes result in the break- down ofincompatibility systems, and autotetraploid forms of self-incompatible species are self-compatible (Lewis, 1942; Levin, 1983). We hypothesize, therefore, that a breakdown or weakening of the incompatibility system accompanied the appearance of tetraploidy in P.pringlei. This may have favored the evolution oftrioecy by initially decreasing the fitness of selfed progeny of hermaphrodites relative to that of the outcrossed seedlings of females. An argument against this hypothesis, however, is the fact that tetraploidy, including autotetraploidy, reduces homozy- gosity in selfed offspring which should lead to reduced inbreeding depression compared to that in selfed offspring of diploids (Murawski et al., in press). Therefore, effects of an increase in ploidy level on the evolution of plant breeding systems are not clear-cut. For example, func- tional dioecy occurs in the tetraploid cactus Echinocereus coccineus whereas the closely related diploid E. triglochidiatus is hermaphroditic (Hoffman, 1992). Trioecy, however, is not associated with tetraploidy in Opuntia robusta (del Castillo, 1986). Furthermore, changes in re- productive success associated with a change in ploidy level have resulted in dioecy in diploids and hermaphroditism or monoecy at higher ploidy levels in Empetrum nigrum and species of Mercurialis, respectively (Guinochet and de Vilmorin, 1975; Durand and Durand, 1992).

Thus far we have discussed the evolution of trioecy based only on models positing nuclear determination of sex. Cytoplasmic genes are known to facilitate the estab- lishment of females and to be involved in many cases of gynodioecy (Ross, 1978; Couvet et al., 1990). Theoretical studies dealing with the evolution of dioecy with nuclear- cytoplasmic sex determination have concluded that the evolution of trioecy is also possible (Gregorius and Ross,

1987; Maurice et al., 1993). The range of relative fertilities ofthe different sex morphs under which trioecy is obtained is broader than in the case of nuclear determination of sex and depends on the details of nuclear-cytoplasmic sex determination. Unfortunately, determining the nature of sex determination in P. pringlei will not be easy, because individuals do not reach sexual maturity until they are approximately 50 years old.

Finally, postulating that the Bahia Kino and cardon forest populations are not in equilibrium with respect to the frequency of the different sexual phenotypes does not explain the presence of unisexuals because our estimates of the relative fertilities of individuals, selfing rates in hermaphrodites, and inbreeding depression do not sup- port a hypothesis of a transitional evolutionary stage to- ward dioecy. Taken at face value, the data suggest the evolution of hermaphroditism from dioecy, but since al- most all cacti, including Pachycereus pecten-aboriginum, a close relative of P. pringlei, are hermaphroditic (Parfitt, 1985; T. H. Fleming, unpublished data), this hypothesis is unlikely to be true.

An alternate explanation for the presence of unisexuals in populations of P. pringlei is that we have underesti- mated levels of inbreeding depression. Not examined here, for example, is the quality of seedlings produced by fe- males and hermaphrodites. Some studies (e.g., Kohn, 1988; Shykoff, 1988; Sakai, Karoly, and Weller, 1989; Agren and Willson, 199 1) indicate that outcrossed seedlings of females have higher relative fitnesses than those of out- crossed and selfed seedlings of hermaphrodites, but others (e.g., del Castillo, 1986) indicate that seedlings of females do not have higher fitness than seedlings of hermaphro- dites. Our data on seedling performance in P. pringlei indicate that seedlings of females and hermaphrodites do not differ significantly in survivorship but do differ sig- nificantly in growth rates under field and laboratory con- ditions; seedlings of hermaphrodites have higher growth rates than those of females (V. Sosa and T. H. Fleming, unpublished data). Hence, all of our data suggest that inbreeding depression is low in this species.

In conclusion, existing models based on nuclear sex determination cannot explain the presence of unisexuals in populations of P. pringlei given our observed values of relative fertilities, selfing, and inbreeding depression at Bahia Kino. Elsewhere we will publish a model based on pollen limitation that appears to provide considerable insight into the evolution of trioecy in P. pringlei (S. Maurice and T. H. Fleming, unpublished data).


AGREN, J., AND M. F. WILLSON. 1991. Gender variation and sexual

differences in reproductive characters and seed production in gyno-

dioecious Geranium maculatum. American Journal of Botany 78:

470480. BOCK, C. 1976. Polymorphisme floral et sttrilite mile chez le Silene acaulis L. Jacq. Comptes Rendus de I'Academie des Sciences de Paris 282: 1865-1868. BUCHMANN,S. L., AND C. W. SHIPMAN. 1990. Pollen harvesting rates

for Apis mellifera L. on Gossypium (Malvaceae) flowers. Journal of

the Kansas Entomological Society 63: 92-100. CHARLESWORTH, 1978.

B., AND D. CHARLESWORTH. A model for the evolution ofdioecy and gynodioecy. American Naturalist 112: 975-


CHARLESWORTH, 198 1. A further study of the problem of the main-

D. tenance of females in gynodioecious species. Heredity 46: 27-39.

, AND B. CHARLESWORTH. 1987. The effect of investment in attractive structures on allocation to male and female functions in plants. Evolution 4 1 : 948-968.

CHARNOV,E. L., J. MAYNARD SMITH,AND J. J. BULL. 1976. Why be an hermaphrodite? Nature 236: 125-1 27. COUVET, D., A. ATLAN, E. BELHASSEN, P. H. GOWON

C. J. GLIDDON, AND F. KJELLBERG. 1990. Coevolution between two symbionts: the case of cytoplasmic male sterility in higher plants. In D. Fu- tuyama and J. Antonovics [eds.], Oxford survey in evolutionary biology, 225-249. Oxford University Press, Oxford.

DARWIN,C. R. 1877. The different forms of flower on plants of the same species. Murray, London.

DELANNAY,X., P. H. GOWON, AND G. VALDEYRON. 1981. Mathematical study of the evolution of gynodioecy with cytoplasmic in- heritance under the effect of a nuclear restorer gene. Genetics 99: 169-181.

DEL CASTILLO,R. F. 1986. La seleccion natural de 10s sistemas de cruzamiento en Opuntia robusta. M.S. thesis, Institucion de En- senanza e Investigacion en Ciencias Agricolas, Centro de Botanica. Chapinga, Mexico.

DOMMEE,B., J. L. BOMPAR, AND N. DENELLE. 1990. Sexual tetra- morphism in Thymelea hirsuta (Thymelaeceae): evidence of the pathway from heterodichogamy to dioecy at the infraspecific level. American Journal of Botany 77: 1449-1462.

DURAND,R., AND B. DURAND. 1992. Dioecie, monoecie, polyploidie et sptciation chez les Mercuriales annuelles. Bulletin de la Socikte botanique de France, Lettres botaniques 139: 377-390.

GOWON, P. H., F. VICHOT,AND J. M. M. VAN DAMME. 199 1. Nuclear- cytoplasmic male sterility: single-point equilibria versus limit cy- cles. American Naturalist 137: 498-5 14.

GREGORIUS,H.-R., AND M. D. ROSS. 1987. Selection with gene-cy- toplasm interactions. 111. Evolution of dioecy. Evolutionary Theory

8: 87-99.

--AND E. M. GILLET. 1983. Selection in plant populations of effectively infinite size. V. Biallelic models of trioecy. Genetics

103: 529-544. GUINOCHETM., AND R. DE VILMORIN.1975. Rore de France, p. 1352. CNRS, Paris.

HOFFMANN,M. T. 1992. Functional dioecy in Echinocereus coccineus (Cactaceae): breeding system, sex ratios, and geographic range of floral dimorphism. American Journal of Botany 79: 1382-1388.

KOHN,J. R. 1988. Why be female? Nature 335: 431433. LEVIN, D. A. 1983. Polyploidy and novelty in flowering plants. American Naturalist 122: 1-25. LEWIS,D. 1941. Male sterility in natural populations of hermaphrodite plants. New Phytologist 40: 56-63. . 1942. The evolution of sex in flowering plants. Cambridge Philosophical Society Biological Reviews 17: 46-6 7. LLOYD, D. G. 1975. The maintenance of gynodioecy and androdioecy in angiosperms. Genetica 45: 1-1 5. , AND K. S. BAWA. 1984. Modification of the gender of seed plants in varying conditions. In M. K. Hecht, B. Wallace, and G.

T. Prance [eds.], Evolutionary biology, vol. 17, 255-338. Plenum, New York, NY.

LOVETTdo us^, J., G. O'BRIEN, AND L. LOVETT DOUST. 1987. Effect of density on secondary sex characteristics and sex ratio in Silene alba (Caryophyllaceae). American Journal of Botany 74: 4046.

MAURICE,S., D. CHARLESWORTH, C. DESFEUX,D. COUVET, AND P. H. GOWON. 1993. The evolution of gender in hermaphrodites of gynodioecious populations with nucleo-cytoplasmic male-sterility. Proceedings of the Royal Society of London, B 25 1: 253-26 1.

MOLAU,U., AND H. C. PRENTICE. 1992. Reproductive system and population structure in three arctic Saxifraga species. Journal of Ecology 80: 149-1 6 1.

MURAWSKI,D. A,, T. H. FLEMING, K. RITLAND, AND J. L. HAMRICK. In press. The mating system of an autrotetraploid cactus, Pachycereus pringlei. Heredity.

PARFITT, B. D. 1985. Dioecy in North American Cactaceae: a review. Sida 1 1: 200-206. Ross, M. D. 1978. The evolution ofgynodioecy and subdioecy. Evolution 32: 174-1 88.


,AND B. S. WEIR. 1976. Maintenance of males and females in hermaphrodite populations and the evolution of dioecy. Evolution

30: 4251141.

SAKAI, A. K., K. KAROLY,AND S. G. WELLER. 1989. Inbreeding de- pression in Schieda globosa and S. salicaria (Caryophyllaceae), sub- dioecious and gynodioecious Hawaiian species. American Journal ofBotany 76: 437-444. , AND S. G. WELLER. 199 1. Ecological aspects of sex expression in subdioecious Schiedea globosa (Caryophyllaceae). American Journal of Botany 78: 1280-1288.

SHREVE,F., AND I. L. WIGGINS. 1964. Vegetation and flora of the Sonoran Desert, vols. 1, 2. Stanford University Press, Stanford, CA.

SHYKOFF, J. A. 1988. Maintenance of gynodioecy in Silene acaulis (Caryophyllaceae): stage-specific fecundity and viability selection. American Journal of Bofany 75: 844-850.

SOLOMON,B. P. 1986. Sexual allocation and andromonoecy: resource investment in male and hermaphrodite flowers of Solanum caro- linense (Solanaceae). American Journal of Botany 73: 12 15-1 22 l.

SUTHERLAND,S. 1986. Patterns of fruit-set: what controls fruit-flower ratios in plants? Evolution 40: 117-128.

TUTTLE,M. D. 199 1. Bats: the cactus connection. National Geograplzic 179(6): 130-140.

YAMPOLSKY, 1922. Distribution of the sex

C. R., AND H. YAMPOLSKY. forms in the phanerogamic flora. Bibliotlzeca Genetica 3: 1-62.

  • Recommend Us