Inbreeding is a major component of the mating system in populations of many plants and animals, particularly hermaphroditic species. In flowering plants, inbreeding can occur through self‐pollination within flowers (autogamy), self‐pollination between flowers on the same plant (geitonogamy), or cross‐pollination between closely related individuals (biparental inbreeding). We performed a floral emasculation experiment in 10 populations of Aquilegia canadensis (Ranunculaceae) and used allozyme markers to estimate the relative contribution of each mode of inbreeding to the mating system. We also examined how these modes of inbreeding were influenced by aspects of population structure and floral morphology and display predicted to affect the mating system. All populations engaged in substantial inbreeding. On average, only 25% of seed was produced by outcrossing (range among populations = 9–37), which correlated positively with both population size (r=+0.61) and density (r=+0.64). Inbreeding occurred through autogamy and biparental inbreeding, and the relative contribution of each was highly variable among populations. Estimates of geitonogamy were not significantly greater than zero in any population. We detected substantial biparental inbreeding (mean = 14% of seeds, range = 4–24%) by estimating apparent selfing in emasculated plants with no opportunity for true selfing. This mode of inbreeding correlated negatively with population size (r= ‐0.87) and positively with canopy cover (r= 0.90), suggesting that population characteristics that increase outcross pollen transfer reduce biparental inbreeding. Autogamy was the largest component of the mating system in all populations (mean = 58%, range = 37–84%) and, as expected, was lowest in populations with the most herkogamous flowers (r= ‐0.59). Although autogamy provides reproductive assurance in natural populations of A. canadensis, it discounts ovules from making superior outcrossed seed. Hence, high autogamy in these populations seems disadvantageous, and therefore it is difficult to explain the extensive variation in herkogamy observed both among and especially within populations.
We studied 10 populations located near the Queen’s University Biological Station in eastern Ontario, Canada (44°35' N, 76°19' W). These populations spanned the range of typical population sizes and densities found in this species, from small, scattered clusters of <100 plants to large, dense patches of >1500 individuals. One population was studied in 1998, another eight in 1999, and one in both years. Because there is substantial turnover of plants between years at any given site (>50% overwinter mortality), and estimates of total selfing and outcrossing do not generally correlate between years among populations (B. Ozimec, C. R. Herlihy, and C. G. Eckert, unpubl. data), we treat the population studied in two years as separate populations in the analyses below. Using only one year of data from this population does not change our results or their interpretation.
Floral Emasculation Experiment
To partition the mating system into its components, we compared self-fertilization between flowers that were emasculated to remove the potential for autogamy and intact control flowers. We randomly selected pairs of plants with the same number of inflorescences and flowers, and matched flowers at the same position in the flowering sequence. The focal flower on one plant in each pair was emasculated by removing the anthers before they began to shed pollen, whereas the focal flower on the other plant was left intact. In each population, we treated about 40 pairs of plants in 1998 and 50–100 pairs in 1999. Emasculation does not damage flowers (Eckert and Schaefer 1998) or reduce their longevity (Griffin et al. 2000; C. R. Herlihy, unpubl. data). We observed pollinators visiting emasculated flowers and did not notice any obvious changes in key aspects of foraging behavior such as the tendency to probe a flower once it had been approached, the probing behavior, or the handling time. Previous studies have also shown that emasculated flowers regularly achieve full seed set or close to it compared to intact flowers, further suggesting that pollinators do not strongly avoid flowers without anthers (Eckert and Schaefer 1998; Herlihy and Eckert 2002). However, the rate of visitation was too low to formally quantify visitation rates or behaviors to intact versus emasculated flowers (e.g. Eckert 2000). Any reduced visitation to emasculated flowers does not change the interpretation of our results because the components of the mating system are estimated as proportions (see below).