1. The effect of variation in floral morphology and display on the level of outcrossing was investigated in a Rocky Mountain population of the self-compatible, protandrous, perennial herb, Aquilegia caerulea. ANOVA revealed substantial variation among 36 plants in floral traits including herkogamy (CV = 35%), protandry (23%), pollen production (12%), flower size (8%), the daily number of flowers in male phase (58%) and autofertility, seed production in the absence of pollinators (192%).
2. Whether this floral variation affected the outcrossing rates of individual plants was assessed using both a categorical analysis, in which outcrossing rates were estimated separately for groups of plants that differed with respect to floral features, and a correlation analysis involving outcrossing rates estimated for individual plants from large progeny arrays (n≥ 40). Both analyses indicated significant positive correlations between the outcrossing rate and herkogamy (r = + 0·71) and protandry (+ 0·54), negative correlations with the number of flowers in male phase during each day of female phase (– 0·42) and autofertility (– 0·36), but no correlation with flower size or pollen production.
3. Partial correlation indicated that both herkogamy and dichogamy, although weakly intercorrelated, were positively correlated with the outcrossing rate and that these effects were only partly mediated by their negative correlation with autofertility.
4. The extent to which floral features and the outcrossing rate covaried across the flowering sequence within plants was also examined in this study. As expected for a plant with strong protandry and sequential blooming, the number of flowers in male phase decreased strongly with flowering sequence position. However, categorical analysis did not support the prediction that the first flowers to open on an inflorescence outcrossed at a lower rate than later-opening flowers.
5. These results suggest that self-fertilization in A. caerulea is the result of all three major modes: autonomous and facilitated autogamy as well as geitonogamy. Although the geitonogamous component of selfing cannot be selected for directly, autogamy may be selected if it provides reproductive assurance.