Phenotypic variation in 11 floral and reproductive traits was studied in cloned plants of Campanula rapunculoides replicated in three discrete environments. Using an ANOVA approach, we determined the relative influence of genotype (G), environment (E), G × E interaction, and ontogeny (position on the raceme) on the 11 traits. Two traits, duration of flowering and pollen size, showed no significant variation. All nine remaining traits had significant genotypic variation, accounting for 21-38% of the total phenotypic variation. Correlations among variant traits in seven genotypes were predominantly positive, but several significant correlations in one environment changed sign or were nonsignificant in another environment. Ovule number was negatively correlated with most male function traits: the negative correlation between ovule and pollen number was particularly strong and consistent across environments. Six traits varied significantly across environments, including number of flowers, number of ovules per flower, and duration of the male phase, but pollen traits did not show a significant environmental main effect. The G × E interaction was significant for flower number, corolla size, nectar quality, duration of the male phase, pollen viability, and ovule number. The contribution of interaction variance to the total phenotypic variation (5-14%) was comparable to that of the environment alone (7-21%). Ovule number decreased in flowers on the upper part of the raceme by nearly 25%, but other traits did not vary significantly by floral position. These results suggest that (1) pollen traits are buffered against environmental change more than ovule number or other floral characters, (2) a male-female trade-off exists and is complicated by ontogenic factors, (3) G × E interactions are common but may have small effects, and (4) specific correlation patterns among floral traits can be dependent upon the environment under which they develop.

The pollen competition hypothesis predicts that when the number of pollen grains deposited onto stigmas exceeds the number of ovules, selection can operate in the time frame between deposition and fertilization. Moreover, because of the overlap in gene expression between the two phases of the life cycle, selection on microgametophytes may alter the resulting sporophytic generation. The extent to which pollen competition occurs in nature has been unclear, because tests of the predictions of the pollen competition hypothesis have used cultivars and/or artificial growth conditions and hand‐pollination techniques. In this study we used a wild species, Cucurbita foetidissima, in its natural habitat (southern New Mexico) to determine the amount and timing of the arrival of pollen onto stigmas, the relationship between pollen deposition and seed number, and the effects of the intensity of pollen competition on progeny vigor. We found that ∼900 pollen grains are necessary for full seed set and that a single visit by a pollinator results in the deposition of 653.0 ± 101.8 pollen grains. About 29% of the flowers receiving a single pollinator visit had 900 or more pollen grains on its stigma. Moreover, within 2 h of anthesis, >4000 pollen grains were deposited onto a typical stigma, indicating that multiple pollinator visits must have occurred. Fruits produced by multiple visits had greater seed numbers (206 vs. 147) than fruits produced by a single visit. Finally, the progeny produced by multiple pollinator visits were more vigorous than those produced by single visits with respect to five measures of vegetative growth (MANCOVA, Wilks’ lambda = 0.96, F6,370 = 2.54, P < 0.02. These data demonstrate that conditions for pollen competition exist in nature and support the prediction that pollen competition enhances offspring vigor.