Ovule number per flower in a world of unpredictable pollination

The number of ovules per flower varies over several orders of magnitude among angiosperms. Here we consider evidence that stochastic uncertainty in pollen receipt and ovule fertilization has been a selective factor in the evolution of ovule number per flower. We hypothesize that stochastic variation in floral mating success creates an advantage to producing many ovules per flower because a plant will often gain more fitness from occasional abundant seed production in randomly successful flowers than it loses in resource commitment to less successful flowers. Greater statistical dispersion in pollination and fertilization among flowers increases the frequency of windfall success, which should increase the strength of selection for greater ovule number per flower. We therefore looked for evidence of a positive relationship between ovule number per flower and the statistical dispersion of pollen receipt or seed number per flower in a comparative analysis involving 187 angiosperm species. We found strong evidence of such a relationship. Our results support the hypothesis that unpredictable variation in mating success at the floral level has been a factor in the evolution of ovule packaging in angiosperms.     

The structure of phenotypic variation in gender and floral traits within and among populations of Spergularia marina (Caryophyllaceae)

We sampled four wild populations of the highly autogamous Spergularia marina (Caryophyllaceae) in California to detect and to measure the magnitude of within- and among-population sources of phenotypic variation in gender and floral traits. From flowers and fruits collected from field and greenhouse-raised plants, we measured ovule number, seed number, mean seed mass, pollen production (greenhouse families only), mean pollen grain volume (greenhouse families only), anther number, anther/ovule ratio, pollen/ovule ratio (estimated using different flowers for pollen than for ovules; greenhouse families only), petal number, and petal size. Using greenhouse-raised genotypes, variation among maternal families nested within populations was evaluated for each trait to determine whether populations differ in the degree of maternally transmitted phenotypic variation. For each population, we used 15 greenhouse-raised maternal families to estimate the broad-sense heritability and genetic coefficient of variation of each floral trait. The magnitude and statistical significance of broad-sense heritability estimates were trait- and population-specific. Each population was characterized by a different combination of floral traits that expressed significant maternally transmitted (presumably genetic) variation under greenhouse conditions. Flowers representing two populations expressed low levels of maternally transmitted variation (three or fewer of nine measured traits exhibited significant maternal family effects on phenotype), while flowers representing the other two populations exhibited significant maternal family effects on phenotype for five or more traits. Our ability to detect statistically significant differences among the four populations depended upon the conditions under which plants were grown (field vs. greenhouse) and on the floral trait observed. Field-collected flowers exhibited significant differences among population means for all traits except anther number. Flowers sampled from greenhouse-raised maternal families differed among populations for all traits except ovule number, seed number, and petal size. We detected negligible evidence that genetic correlations constrain selection on floral traits in Spergularia marina.     

The evolution of ovule number and flower size in wind-pollinated plants

In angiosperms, ovules are "packaged" within individual flowers, and an optimal strategy should occur depending on pollination and resource conditions. In animal-pollinated species, wide variation in ovule number per flower occurs, and this contrasts with wind-pollinated plants, where most species possess uniovulate flowers. This pattern is usually explained as an adaptive response to low pollen receipt in wind-pollinated species. Here, we develop a phenotypic model for the evolution of ovule number per flower that incorporates the aerodynamics of pollen capture and a fixed resource pool for provisioning of flowers, ovules, and seeds. Our results challenge the prevailing explanation for the association between uniovulate flowers and wind pollination. We demonstrate that when flowers are small and inexpensive, as they are in wind-pollinated species, ovule number should be minimized and lower than the average number of pollen tubes per style, even under stochastic pollination and fertilization regimes. The model predicts that plants benefit from producing many small inexpensive flowers, even though some flowers capture too few pollen grains to fertilize their ovules. Wind-pollinated plants with numerous flowers distributed throughout the inflorescence, each with a single ovule or a few ovules, sample more of the airstream, and this should maximize pollen capture and seed production.     

Sources of variation in pollinator contribution within a guild: the effects of plant and pollinator factors

Among plants visited by many pollinator species, the relative contribution of each pollinator to plant reproduction is determined by variation in both pollinator and plant traits. Here we evaluate how pollinator movement among plants, apparent pollen carryover, ovule number, resource limitation of seed set, and pollen output affect variation in contribution of individual pollinator species to seed set in Lithophragma parviflorum (Saxifragaceae), a species visited by a broad spectrum of visitors, including beeflies, bees and a moth species. A previous study demonstrated differences among visitor species in their single-visit pollination efficacy but did not evaluate how differences in visitation patterns and pollen carryover affect pollinator efficacy. Incorporation of differential visitation patterns and pollen carryover effects —commonly cited as potentially important in evaluating pollinator guilds — had minor effects (0–0.6% change) on the estimates of relative contribution based on visit frequency and single-visit efficacy alone. Beeflies visited significantly more flowers per inflorescence than the bees and the moth. Seed set remained virtually constant during the first three visited flowers for beeflies and larger bees, indicating that apparent pollen carryover did not reduce per-visit efficacy of these taxa. In contrast, Greya moth visits showed a decrease in seed set by 55.4% and the smaller bees by 45.4% from first to second flower. The larger carryover effects in smaller bees and Greya were diminished in importance by their small overall contribution to seed set. Three variable plant traits may affect seed set: ovule number, resource limitation on seed maturation, and pollen output. Ovule number per flower declined strongly with later position within inflorescences. Numbers were much higher in first-year greenhouse-grown plants than in field populations, and differences increased during 3 years of study. Mean pollen count by position varied 7-fold among flowers; it paralleled ovule number variation, resulting in a relatively stable pollen:ovule ratio. Resource limitation of seed set increased strongly with later flowering, with seed set in hand-pollinated flowers ranging from 66% in early flowers to 0% in the last two flowers of all plants. Variation in ovule number and resource limitation of seed maturation jointly had a strong effect on the number of seeds per flower. Visitation to early flowers had the potential to cause more seed set than visitation to later flowers. Overall, the most important sources of variation to seed production contribution were differences among pollinators in abundance and absolute efficacy (ovules fertilized on a single visit) and potentially differential phenology among visitor species. These effects are likely to vary among populations and years.     

Efficiency and allometry of seed production in grassland forbs: site and year comparative study in the Italian Eastern Pre-Alps

Forbs are important biodiversity components of grasslands and are often threatened by management intensification. As most forbs propagate predominantly by seed, knowledge of their seed regenerative traits would improve the conservation and restoration of forb-rich ecosystems. The main seed production traits of six forbs that are common in European species-rich grasslands were studied by collecting fertile shoots from different sites and over several years. Among sites and years, variability was high, particularly in the number of inflorescences per shoot, which affected ovule production more than any other trait. Relationships between inflorescence size and the number of ovules were mainly negatively allometric or almost so, with lower flower densities in larger inflorescences. The average ovule-to-seed transformation efficiency was 58%. There was significant variation among collections of the same species, and even more between species. Species with a low ovule-to-seed transformation efficiency generally exhibited compensatory, high seed viability. Large inflorescences had high ovule to seed utilization values, probably because of better nutrient conditions. Seed germinability (average, 30%) was much lower than seed viability (average, 54%); therefore, seed dormancy was an important feature of the species studied.     

Structural aspects of ovule and seed development and nonrandom abortion inMelilotus officinalis (Fabaceae)

Summary Only one ovule matures into a seed inMelilotus officinalis. Although eight ovules form within an ovary, only the basal ovule develops into a mature seed, whereas the other ovules degenerate. The investigation of ovule and seed structure at different developmental stages and a comparison of quantitative characters of differently fated ovules within an ovary were undertaken by light, phase contrast, and fluorescence microscopy. In this species, campylotropous ovules develop simultaneously on marginal placentae in an apocarpous unilocular gynoecium. Megasporo- and megagametogenesis proceed normally and are completed in bud. The maturation of the Polygonum type embryo sac takes place after the flower opens. Shortly before fertilization, synergids show signs of degeneration in all ovules. At this stage, neither the structure nor the sizes of ovules within one ovary differ significantly. In spite of this, only the basal ovule develops into a seed. Rarely, one of the upper-situated ovules or the basal and another ovule mature into seeds. Seed enlargement is insignificant until the stage when globular embryo and nuclear endosperm are formed. At the seed-filling stage, other ovules have collapsed and the seed gradually comes to occupy the total volume of the pod. The fruit-to-seed length ratio decreases considerably during seed ripening. At fertilization, ovary length is four times greater than ovule length. In the mature state, the fruit and seed lengths are approximately equal. Seed size and weight diminish with an increase in seed number within a pod, although pod size remains constant. It is assumed that nonrandom abortion of young seeds inM. officinalis is under maternal control and is not related to structural abnormalities in ovule development or with limitation in pollen. We suppose that evolution of this species may have proceeded in the direction of a decrease in seed number and an increase in its sizes, which may play an important role in seed dispersal and seedling establishment.     

Patterns of fruit and seed production inBauhinia ungulata (Leguminosae)

Patterns of seed and fruit production ofBauhinia ungulata, a small tree legume indigenous in tropical America, were studied in Costa Rica. Only about 8% of flowers produced fruits. The average pod had 19 ovules and about two thirds of these began seed development, with mature pods containing an average of 9.7 mature undamaged seeds. About half of the mature pods were damaged by herbivores and within these, 27% of ovules or seeds had been eaten. Among trees there was no significant variation in pod production, but the number of ovules per pod and seed production per pod varied significantly. Within infructescences most pods were retained at middle positions. Within pods, the probability of an ovule developing into a seed increased toward the distal end. The pattern of seed and fruit production in this species agrees well in general with that reported for other neotropical legumes. The abortion of seeds and fruits can be regarded as a way of controlling maternal investment, and as a response to herbivory.     

Factors limiting seed production of Taxus brevifolia (Taxaceae) in Western Oregon

Seed production of Pacific yew (Taxus brevifolia), an understory conifer, was studied at four sites in western Oregon over 2 yr. The effects of pollen supplementation, overstory canopy, and predator exclusion on ovule attrition were examined. Supplemental hand-pollination of ten trees at two sites resulted in significantly increased rates of ovule development and a doubling of seed efficiency (ratio of seeds to ovules). However, seed efficiency still averaged <15% on branches receiving supplemental pollen, so pollination was not a primary factor limiting seed production. The number of developing ovules was positively associated with overstory openness, but seed production was not. Seed efficiency was negatively associated with overstory openness. Branches bagged to exclude vertebrate seed predators had higher seed production than unbagged branches at three of four sites for 2 yr. In contrast to unbagged branches, seed production on bagged branches was positively associated with overstory openness, as was the effectiveness of bagging. Therefore, both vertebrate predation and overstory were important in limiting seed production, and these factors interacted. Factors limiting seed production varied in importance among the four sites and between years, illustrating the importance of examining multiple limiting factors over several sites and years.     

POLLEN-OVULE RATIOS AND BREEDING SYSTEM EVOLUTION IN SOLANUM SECTION BASARTHRUM (SOLANACEAE)

Pollen-ovule ratio, pollen grains per flower, pollen grain volume, pollen biomass, ovules per ovary, seeds per fruit, seed volume, seed biomass, and corolla size were studied to reveal patterns of variation within a monophyletic group. Solanum sect. Basarthrum includes 22 neotropical species of two clades, one having small, few-seeded red fruits, and the other having larger, many-seeded green fruits. The former includes self-incompatible species and a dioecious species, while the latter includes self-incompatible species, self-compatible (and autogamous) species, and a domesticate (the “pepino”). Although the pollen quantity and ovule quantity of the self-incompatible species are significantly higher in the green-fruited subgroup than in the red-fruited subgroup, the pollen-ovule ratios of the self-incompatible species are not significantly different between the two subgroups, suggesting parallel evolution of the pollen-ovule ratio. Furthermore, the pollen-ovule ratio tracks the breeding system: self-incompatible species have significantly higher pollen-ovule ratios than self-compatible species, resulting both from fewer pollen grains and more ovules of the latter. The pollen-ovule ratio of the dioecious species is among the highest of all, resulting only from fewer ovules. The pepino is self-compatible but has a pollen-ovule ratio like its wild self-incompatible progenitors and shows wide variation in seed production and pollen quality, but not in pollen quantity. Among all species, pollen size and quantity are inversely proportional, as are seed size and quantity.     

Size-dependent sex allocation within flowers of the annual herb Clarkia unguiculata (Onagraceae): ontogenetic and among-plant variation

The relative allocation of resources to male and female functions may vary among flowers within and among individual plants for many reasons. Several theoretical models of sex allocation in plants predict a positive correlation between the resource status of a flower or individual and the proportion of reproductive resources allocated to female function. These models assume that, independent of resource status, a negative correlation exists between male and female investment. Focusing on the allocation of resources within flowers, we tested these theoretical predictions and this assumption using the annual Clarkia unguiculata (Onagraceae). We also sought preliminary evidence for a genetic component to these relationships. From 116 greenhouse-cultivated plants representing 30 field-collected maternal families, multiple flowers and fruits per plant were sampled for gamete production, pollen?:?ovule ratio, seed number, ovule abortion, seed biomass/fruit, mean individual seed mass, and petal area. If sex allocation changes as predicted, then (1) assuming that flowers produced early have access to more resources than those produced later, basal flowers should exhibit a higher absolute and proportional investment in female function than distal flowers and (2) plants of high resource status (large plants) should produce flowers with a higher proportional investment in female function than those of low resource status. Within plants, variation in floral traits conformed to the first prediction. Among plants and families, no significant effects of plant size (dry stem biomass) on intrafloral proportional sex allocation were observed. We detected no evidence for a negative genetic correlation between male and female investment per flower, even when controlling for plant size.     

The interplay of pollination,costs of reproduction and plant size in maternal fertility limitation in perennial <Emphasis Type="Italic">Paeonia officinalis</Emphasis>

Although several factors can limit female fertility in perennial plants, rarely have they been jointly studied in a single species over several years. In this study we experimentally manipulate seed production and simultaneously analyse the potential contribution of pollen limitation, costs of reproduction and plant size to variation in seed output over a 3-year period in the perennial herb Paeonia officinalis, in southern France. Since this rare species is threatened by forest closure in many sites we also examine the causes of female fertility variation in relation to habitat closure (open habitat vs. woodland). P. officinalis has a partial self-incompatibility system and only very low ability for autonomous self pollination in the absence of pollinators. However, supplementary pollination of individual plants in three consecutive years did not significantly increase seed production above natural levels. Forest closure was associated with a decline in ovule and seed production, which again was not due to pollen limitation since supplementary pollination had no significant effect on seed set in the woodland habitat. Comparison of the maternal fertility of plants which were previously excluded from reproduction with those which were hand pollinated to maximise seed set in two previous years produced no evidence that seed production in year three is limited by costs associated with prior reproduction. Likewise, flowering probability was not related to prior seed production but was however positively related to plant size. The absence of any influence of pollen limitation or prior reproduction on seed production suggests that sub-maximal seed production in long-lived perennial herbs may be part of a size-dependent strategy that maximises life-time seed production and fitness without compromising survival.     

Size-dependent gender modification in a hermaphroditic perennial herb

The size-advantage model predicts that hermaphroditic organisms adjust sex allocation depending on their resource status. We investigated the relationship between size and sex allocation in the co-sexual perennial herbs Trillium erectum and Trillium grandiflorum at two sites in southern Ontario, Canada by measuring pollen and ovule production and biomass allocation at flowering and fruiting. In both species, there was a strong relationship between size and gender; larger plants allocated proportionately more biomass to female reproduction and produced fewer pollen grains relative to ovules than smaller plants. Variation in gender was better explained by size than age, although age and size were correlated. While the relationship between size and gender was similar between species, T. erectum allocated proportionately more to female reproduction than T. grandiflorum, independent of size. In the absence of pollen limitation, there was no evidence of secondary adjustment of gender at fruiting. The results are discussed in the context of models predicting size-dependent gender modification in animal-pollinated plants. Evidence about the pollination and seed dispersal biology of Trillium spp. suggests that the relative effects of local mate and resource competition may be important in driving size-dependent sex allocation in these species.     

Floral display, pollinator visitation and reproductive success in the dioecious perennial herb Wurmbea dioica (Liliaceae)

Floral traits that increase attractiveness to pollinators are predicted to evolve through selection on male function rather than on female function. To determine the importance of male-biased selection in dioecious Wurmbea dioica, we examined sexual dimorphism in flower size and number and the effects of these traits on pollinator visitation and reproductive success of male and female plants. Males produced more and larger flowers than did females. Bees and butterflies responded to this dimorphism and visited males more frequently than females, although flies did not differentiate between the sexes. Within sexes, insect pollinators made more visits to and visited more flowers on plants with many flowers. However, visits per flower did not vary with flower number, indicating that visitation was proportional to the number of flowers per plant. When flower number was experimentally held constant, visitation increased with flower size under sunny but not overcast conditions. Flower size but not number affected pollen removal per flower in males and deposition in females. In males, pollen removal increased with flower size 3 days after flowers opened, but not after 6 days when 98% of pollen was removed. Males with larger flowers therefore, may have higher fitness not because pollen removal is more complete, but because pollen is removed more rapidly providing opportunities to pre-empt ovules. In females, pollen deposition increased with flower size 3 days but not 6 days after flowers opened. At both times, deposition exceeded ovule production by four-fold or more, and for 2 years seed production was not limited by pollen. Flower size had no effect on seed production per plant and was negatively related to percent seed set, implying a tradeoff between allocation to attraction and reproductive success. This indicates that larger flower size in females is unlikely to increase fitness. In both sexes, gamete production was positively correlated with flower size. In males, greater pollen production would increase the advantage of large flowers, but in females more ovules may represent a resource cost. Selection to increase flower size and number in W. dioica has probably occurred through male rather than female function. Received: 15 June 1997 / Accepted: 12 February 1998     

鹅掌楸的传粉环境与性配置

近年来研究表明,动物传粉者对植物花部诱物特征（花冠形状大小,花蜜产量和花序大小）有潜在的选择作用。不同效率的传粉者可能导致植物性配置的变化,不同传粉效率的环境下,两性花植物鹅掌楸的花粉和胚珠的配置不同,居群Ｚ在主要访花者是传粉效率较低的类的传粉环境下,花粉粒小,花粉数量相对较高;另一方面胚珠投资的减少,缓解了低效的传粉（增大授粉的几率）,增加了受精的机会。相对而言,具有高效访花者的两个居群,资源较     

ECOLOGICAL CONSEQUENCES AND PHENOTYPIC CORRELATES OF PETAL SIZE VARIATION IN WILD RADISH,RAPHANUS SATIVUS (BRASSICACEAE)

In this paper we examine some ecological consequences and phenotypic correlates of flower size variation in wild radish, Raphanus sativus. Mean corolla diameter varied significantly among individuals within natural populations of R. sativus in California. On the average, almost 40% of flower biomass was allocated to corolla tissue. In field experiments, pollinator visitation increased significantly with corolla size. Large flowers also accumulated more nectar when pollinators were excluded from plants. In three populations, corolla size was positively correlated with allocation to pollen per flower (either anther weight or pollen grain number), but there was usually no phenotypic relationship between corolla size and several measures of female allocation (ovule number per flower, proportion fruit set, and total seed mass per fruit). Plants growing in the field produced fewer large flowers per unit of stem, and stem biomass was negatively related to corolla size for plants grown under controlled greenhouse conditions. Male and female fitness may covary differently with allocation to attractive floral features in species such as R. sativus, where seed production is often limited by resources rather than by pollen.     

Self pollination and resource availability affect ovule abortion inCassia fasciculata (Caesalpiniaceae)

We examined the effects of pollen source and resource availability on ovule abortion in the annual legumeCassia fasciculata. Pollen source was controlled by hand-pollinating flowers with cross- or self-pollen. Resource availability to developing fruits was controlled by adjusting fruit loads (heavy versus light) on each plant and exposing plants to different photoperiod cycles (16 vs 12 h of light; short days favor fruit growth at the expense of vegetative growth). In mature fruits the proportion of ovules expanding (showing some development over virgin ovules) ranged from 89–95% and did not increase with resource availability, suggesting that unexpanded ovules were either unfertilized or obligately aborted shortly after fertilization. The proportion of expanded ovules maturing in mature fruits was near 97% for both self- and crosspollinations in the treatment with highest resource availability (light load, short days) and lower in the remaining treatments, where self-pollination resulted in up to 9% lower seed maturation than cross-pollination. In the latter three treatments a pollen source effect was dependent upon the maternal plant; in some plants selfing increased abortion and in others it did not. Collectively, the results suggest that (1) both pollen source and resource availability affect ovule abortion, (2) at least some abortion is facultative, and (3) when resources are limited, self-pollination increases abortion in some but not all maternal plants.     

Environmental contribution to floral trait variation in Chamaecrista fasciculata (Fabaceae: Caesalpinoideae)

Although intraspecific variation in plant floral traits has been documented for a number of plant species, the causes of such variation are largely unknown. We first quantified floral trait variation in an Illinois prairie population of Chamaecrista fasciculata Michx. We then used a field experiment to determine the contribution of leaf herbivory to this variation and a greenhouse experiment to determine the contribution of leaf herbivory, and variable soil nutrient and water content to floral trait variation. Variation in environmental factors explained a significant portion of the naturally occurring variation in corolla width, ovule number, ovule size, and anther length. In the field, manual removal of 25% or more leaf area reduced ovule size and anther length (and by inference, pollen production), and delayed flowering. In the greenhouse, plants from which we removed 25% or more of their leaf area or which were given limited water produced fewer ovules than control plants. Addition of nutrients interacted with soil moisture to affect corolla diameter and ovule number. Despite our demonstration of significant environmental impacts on reproductive traits, these impacts were relatively much smaller than those on plant size, suggesting that floral traits are buffered against variable resource availability.     

Size-dependent sex allocation in hermaphroditic plants: the effects of resource pool and self-incompatibility

The effects of the resource pool and resource obtained during a season for seed maturation and self-incompatibility on the size-dependency of evolutionarily stable sex allocation were analysed theoretically. In hermaphroditic plants, reproductive resources allocated between male and female function may not be paid from a single resource pool, because plants can mature seeds using not only reserved resources but also newly gained resources after flowering. But the resource investment to male function is limited to the flowering stage. Under the assumption of constant reserve efficiency and diminishing resource return per investment to leaves, large plants should use both reserved and newly gained resources for seed maturation, while small plants should use only new resources. When both reserved and new resources are used, the optimal allocation for self-compatible species is to invest a constant amount of resources into male function irrespective of resource size, because the female fitness curve increases linearly and the male curve decelerates due to local mate competition. In self-incompatible species, on the other hand, fitness gain per investment through male function and the optimal amount of resources invested in male function decrease with size. Thus a decrease in maleness with size should be emphasized more in self-incompatible species than in self-compatible one. When only new resources are used for seed growth, the female fitness curve as well as male one decelerates with investment. Consequently, the investment in both male and female functions should increase with size, in both self-compatible and self-incompatible species. The magnitude of reserve efficiency relative to efficiency of resource gain after flowering affects size-dependent pattern of sex allocation, while the cost of seed maturation relative to ovule production has little effect on it. The plant size variation in a population emphasizes size-dependency of sex allocation. When size variation is large enough, it is possible that large plants become complete female in self-incompatible species, but it is not in self-compatible species.     

Effects of floral sexual investment and dichogamy on floral longevity

Aims Floral longevity, the duration that a flower remains open and functional, varies greatly among species. Variation in floral longevity has been considered to be optimal strategy for resource allocation under different ecological conditions, mainly determined by the rates of pollination and cost of flower maintenance. However, it is unclear whether an intrinsic factor, floral sexual investment, constrains evolution of floral longevity. The theoretical model also predicts that dichogamy favors long-lived flowers, but empirical studies to test this prediction remain unexplored.Methods To examine the effect of floral sexual investment on floral longevity, we measured flower size together with pollen and ovule production in 37 sympatric flowering plants in a natural community. The duration of the female and male phase in 21 protandrous species and floral longevity of the other 16 adichogamous species were documented in the field.Important findings Floral longevity varied from 1 day to 15 days, while pollen number per flower varied from 643 to 710880 and ovule number per flower from 1 to 426 in the 37 species. Flower size was correlated with pollen production as well as ovule production. Floral longevity was positively related to pollen production but not to ovule production. Consistent with the prediction that dichogamy favors long-lived flowers, we found the floral longevity of protandrous species was significantly longer than that of adichogamous species. In the protandrous species, pollen production per flower was observed to be positively related to male duration, while ovule production was not related to female duration. Our analyses of variation in floral longevity and sexual investment among different species suggest that the floral sexual investment could be an intrinsic factor contributing to the selected floral longevity, particularly the male phase, and that high pollen production could potentially increase pollen removal, i.e. male productive success.     

RESPONSES OF FLORAL TRAITS TO SELECTION ON PRIMARY SEXUAL INVESTMENT IN SPERGULARIA MARINA: THE BATTLE BETWEEN THE SEXES

Two widespread assumptions underlie theoretical models of the evolution of sex allocation in hermaphroditic species: (1) resource allocations to male and female function are heritable; and (2) there is an intrinsic, genetically based negative correlation between male and female reproductive function. These assumptions have not been adequately tested in wild species, although a few studies have detected either genetic variation in pollen and ovule production per flower or evidence of trade-offs between male and female investment at the whole plant level. It may also be argued, however, that in highly autogamous, perfect-flowered plant taxa that exhibit genetic variation in gamete production, strong stabilizing selection for an efficient pollen:ovule ratio should result in a positive correlation among genotypes with respect to mean ovule and mean pollen production per flower. Here we report the results of a three-generation artificial selection experiment conducted on a greenhouse population of the autogamous annual plant Spergularia marina. Starting with a base population of 1200 individuals, we conducted intense mass selection for two generations, creating four selected lines (high and low ovule production per flower; high and low anther production per flower) and a control line. By examining the direct and correlated responses of several floral traits to selection on gamete production per flower, we evaluated the expectations that primary sexual investment would exhibit heritable variation and that resource-sharing, variation in resource-garnering ability, or developmental constraints mold the genetic correlations expressed among floral organs. The observed direct and correlated responses to selection on male and female gamete production revealed significant heritabilities of both ovule and anther production per flower and a significant negative genetic correlation between them. When plants were selected for increased ovules per flower over two generations, ovule production increased and anther production declined relative to the control line. Among plants selected for decreased anthers per flower, we observed a decline in anther production and an increase in ovule production relative to the control line. In contrast, the lines selected for low ovules per flower and for high anthers per flower exhibited no evidence for significant genetic correlations between male and female primary investment. Correlated responses to selection also indicate a genetically based negative correlation between the production of normal versus developmentally abnormal anthers (staminoid organs); a positive correlation between the production of ovules versus staminoid organs; and a positive correlation between the production of anthers and petals. The negative relationship between male versus female primary investment supports classical sex allocation theory, although the asymmetrical correlated responses to selection indicate that this relationship is not always expressed.