Localized dispersal and mating may genetically structure plant populations, resulting in matings among related individuals. This biparental inbreeding has significant consequences for the evolution of mating systems, yet is difficult to estimate in natural populations. We estimated biparental inbreeding in two populations of the largely self‐fertilizing plant Aquilegia canadensis using standard inference as well as a novel experiment comparing apparent selfing between plants that were randomly relocated within populations to experimental control plants. Using two allozyme markers, biparental inbreeding (b) inferred from the difference between single‐locus and multilocus estimates of selfing (b=ss–sm) was low. Less than 3% of matings involved close relatives (mean b= 0.029). In contrast, randomly relocating plants greatly reduced apparent selfing (mean ss= 0.674) compared to control plants that had been dug up and replanted in their original locations (ss= 0.953, P= 0.002). Based on this difference in ss, we estimated that approximately 30% of all matings involved close relatives (mean b= 0.279, 95% CL = 0.072–0.428). Inference from ss–sm underestimated b in these populations by more than an order of magnitude. Biparental inbreeding is thought to influence the evolution of self‐fertilization primarily through reducing the genetic cost of outcrossing. This is unlikely to be of much significance in A. canadensis because inbreeding depression (a major cost of selfing) is much stronger than the cost of outcrossing. However, biparental inbreeding combined with strong inbreeding depression may influence selection on dispersal.
We conducted this experiment during spring of 2000 in two populations located within 15 km of the Queen’s University Biological Station near Chaffey’s Lock in eastern Ontario, Canada (populations QFP1 and QOR1). These populations were typical in that plants were scattered in clumps in shallow soil around the margins of granite outcrops. Both contained many reproductive individuals that survived to produce fruit (QFP1, about 250; QOR1, about 1500), so that moving modest samples of plants would have little effect on spatial genetic structure. However, the two populations differed in area and plant density. QFP1 consisted of plants scattered over a 75 X 50 m area. The distance between neighboring reproductive individuals averaged (±SE) 81 ± 7 cm (n = 87 neighbor pairs), and the number of reproductive plants in a 1-m radius around focal plants averaged 1.3 ± 0.1. In contrast, QOR1 was a dense population occupying a 15 X 10 m area (nearest-neighbor distance = 16 ± 2 cm, number of plants in 1-m radius = 27.8 ± 2.4, n = 75). In each population, we carefully dug up about 100 plants that had just initiated buds and randomly assigned them to two treatments: half were replanted in their original locations (control; C) and the other half were moved to a randomly chosen position that was previously occupied by another moved plant and was >3 m from its original location (moved; M). Transplanting occurred before flowering. Samples decreased to 17 M and 14 C plants in QFP1 and 14 M and 13 C plants in QOR1 through mortality caused by drought and heavy herbivory that typically occurs in populations of A. canadensis in eastern Ontario (Herlihy and Eckert 2002; A. Kliber and C. G. Eckert, unpubl. data). We collected one mature fruit per surviving experimental plant as well as from a random sample of about 50 unmanipulated plants per population and screened 10 seeds/fruit for two diallelic polymorphic allozyme loci (isocitric dehydrogenase, IDH: EC 220.127.116.11; and peroxidase, PER: EC 18.104.22.168) following Routley et al. (1999). The frequency of the slow-migrating allele was 0.552 and 0.790 for IDH and 0.248 and 0.217 for PER in QFP1 and QOR1, respectively. The proportion of seeds produced through selfing was estimated using the mixed mating model (Ritland 1986) as implemented in the maximumlikelihood computer program MLTR (Ritland 1990a).