DOES SELF-POLLINATION PROVIDE REPRODUCTIVE

Transcription

1 American Journal of Botany 85(7): DOES SELF-POLLINATION PROVIDE REPRODUCTIVE ASSURANCE IN AQUILEGIA CANADENSIS (RANUNCULACEAE)? 1 CHRISTOPHER G. ECKERT 2 AND AMY SCHAEFER Department of Biology, Queen s University, Kingston, Ontario, Canada K7L 3N6 The ability to produce seeds when pollinators or potential mates are scarce is thought to be one of the main advantages of self-fertilization in flowering plants. However, whether autonomous selfing increases seed set in natural populations has seldom been tested, and even fewer studies have evaluated the advantage of selfing across a gradient of pollen availability. This study examines the fertility consequences of autonomous selfing in Aquilegia canadensis (Ranunculaceae), a shortlived, spring-flowering perennial typically found in small, patchy populations on rock outcrops. We used a pollinator exclusion experiment to confirm reports that A. canadensis has a well-developed capacity for autonomous selfing resulting from incomplete protogyny and close proximity of stigmas and anthers during dehiscence. Flowers excluded from pollinators set 87% as many seeds per carpel (X 1SE seeds) as hand-pollinated flowers ( seeds), and seed production in unpollinated flowers correlated negatively with the distance between stigmas and anthers (r 0.46). Autonomous selfing could be potentially valuable in providing reproductive assurance because only pollen grains were deposited on each stigma before anther dehiscence, compared to pollen grains by the end of anther dehiscence. However, prevention of autonomous selfing by anther removal before dehiscence did not decrease seed set, even for plants at low plant densities where outcross pollen may have been in short supply. Emasculated flowers set as many seeds per carpel ( ) as intact flowers ( ), indicating that sufficient cross pollen is deposited for full seed set. These results do not support the hypothesis that autonomous selfing by A. canadensis has been selected because it provides reproductive assurance. Key words: Aquilegia canadensis; autogamy; delayed self-fertilization; dichogamy; floral morphology; herkogamy; plant mating systems; protogyny; Ranunculaceae. The ability to produce seeds when pollinators or potential mates are scarce is proposed as a principal advantage of self-fertilization in flowering plants (Darwin, 1876; Baker, 1955; Jain, 1976; Lloyd, 1980; Holsinger, 1996). Selection for the reproductive assurance provided by selfing may be especially important in explaining the widespread occurrence of partial selfing in species where the fitness value of producing selfed progeny is greatly reduced by inbreeding depression (Cruden and Lyon, 1989; Schoen, Morgan, and Bataillon, 1996). Despite the widely accepted importance of reproductive assurance, there has been very little investigation of whether the capacity for self-fertilization increases seed set in natural populations. Most support for the selective value of reproductive assurance is based on loose correlations between breeding system and ecological circumstance. The most commonly cited evidence is that, compared to obligate outcrossers, species capable of self-fertilization tend to have wider geographical ranges, occur in marginal or temporary habitats, or have undergone major colonization episodes in 1 Manuscript received 8 July 1997; revision accepted 17 November The authors thank Floyd Connor, Marcel Dorken, Amy Macdougall, Katherine Mavraganis, Grace Lo, and Steve Lougheed for help with field work; the Queen s University Biological Station (QUBS) for field station facilities; Bob Montgomerie for helpful suggestions; Lawrence Harder, Kent Holsinger, Katherine Mavraganis, John Pannell, and Matthew Routley for comments on the manuscript; and the National Sciences and Engineering Research Council of Canada (NSERC) for a research grant to CGE. 2 Author for correspondence (phone: , FAX: , eckertc@biology.queensu.ca). 919 their recent evolutionary history (reviewed in Lloyd, 1980). For example, self-compatible taxa are more likely to establish after long-distance dispersal to oceanic islands than self-incompatible taxa (Baker, 1955, 1967; Cox, 1989). Additional support comes from observations that many species produce seeds through autonomous self-pollination when isolated from pollinators in a greenhouse or pollinator exclusion cage (Lloyd and Schoen, 1992) and that species that readily self-pollinate autonomously often occur in ecological situations where pollinators tend to be scarce (e.g., Hagerup, 1951; Garnock- Jones, 1976; Schemske, 1978; Motten, 1982; Dole, 1990, 1992). Furthermore, within some species the capacity for autonomous selfing negatively correlates with pollinator service among populations (e.g., Ganders, 1974; Wyatt, 1986; Barrett, Morgan, and Husband, 1989; Rathcke and Real, 1993; Inoue, Maki, and Masuda, 1996). Although these correlations are consistent with a role for reproductive assurance in the evolution of self-fertilization, they are usually complicated by correlations between factors that affect the level of cross-pollination (pollinator abundance, mate availability) and those that facilitate the selection of selfing for other reasons (e.g., small population size, founder effect, unsaturated habitats; Lloyd, 1980). Direct experimental tests of whether autonomous selfing increases seed set in natural populations are few (Bernhardt, 1976; Motten, 1982; Piper, Charlesworth, and Charlesworth, 1986; Cruden and Lyon, 1989; Klips and Snow, 1997). Reproductive assurance may potentially be afforded by both autonomous autogamy (automatic within-flower selfing) and, to a lesser extent, facilitated autogamy (within-

2 920 AMERICAN JOURNAL OF BOTANY [Vol. 85 flower selfing caused by insect visitation; Lloyd and Schoen, 1992). Whether either mode of autogamy increases seed production can be determined experimentally by removing anthers from flowers before dehiscence (Schoen and Lloyd, 1992). If autogamy provides reproductive assurance, then emasculated flowers should set fewer seeds than intact flowers. Because the contribution of autonomous autogamy to seed set depends on pollinator abundance and/or potential mates, reproductive assurance should be context dependent. Emasculations should, ideally, be performed across a gradient of pollen availability. However most experimental studies of reproductive assurance have involved only a single population during one flowering season, with no attempt to examine the effect of pollen availability on the fertility value of autonomous selfing (but see Motten, 1982; Piper, Charlesworth, and Charlesworth, 1986). Here, we experimentally test whether the capacity for autogamous self-fertilization provides reproductive assurance in a columbine, Aquilegia canadensis L. (Ranunculaceae). This species is a short-lived, spring-flowering perennial of rocky outcrops, cliffs, and dry woods in eastern central North America (Payson, 1918; Munz, 1946). Like all its congeners, flowers consist of several separate carpels (4 5) surrounded by a column of many stamens (30 40), with petals modified in the form of nectar spurs. The nodding flowers are primarily red, with some yellow on the widened entrances (laminae) to the spurs. Red coloration combined with the production of dilute sugary nectar (25% sucrose; Macior, 1978) suggests adaptation to hummingbird pollination (Grant and Grant, 1968; Grant and Temeles, 1992), though several species of bumble bees (Bombus spp.) also appear to pollinate effectively as well as rob nectar (Macior, 1966). Although the showy, rewarding flowers appear adapted for cross-pollination, A. canadensis can produce abundant seeds when isolated from pollinators (Macior, 1978; K. Mavraganis and C.G. Eckert, Queen s University, unpublished data). This capacity for autonomous seed production could be valuable, because A. canadensis may often receive insufficient cross pollen due to low plant density in its patchy habitat and reduced pollinator visitation during inclement spring weather. The mechanism underlying this autonomous seed production involves incomplete protogyny combined with close proximity of stigmas and anthers during dehiscence. Early during floral development, just after the nectar spurs have started to develop and the corolla has begun to turn red, the styles elongate so that stigmas are presented well beyond the enclosing sepals. The flower then continues to enlarge and the sepals spread to reveal the openings of the nectar spurs. About 5 d after the stigmas are first presented, the first group of 5 10 stamens straighten out to place dehiscing anthers within 3 mm of the stigmas. During the next 3 d, all stamens straighten and dehisce. These observations suggest that incomplete protogyny could cause delayed self-pollination in A. canadensis. However, the extent to which delayed self-pollination provides reproductive assurance depends on how many outcross pollen grains are deposited on stigmas before anther dehiscence. If stigmas are saturated with outcross pollen before anthers start dehiscing and fertilization follows quickly, then later-arriving self-pollen cannot contribute to seed set. Similarly, the extent to which self-pollen is required for full seed set also depends on how many outcross pollen grains are deposited on stigmas during anther dehiscence. In this study, we used pollinator exclusion and handpollinations both to quantify the seed set afforded by autonomous self-fertilization and to identify the floral features that promote autonomous selfing. We then examined the arrival of pollen grains before and after anther dehiscence to determine the potential contribution of selfpollination to seed set. Finally, we performed an emasculation experiment to quantify the actual contribution of autogamous pollination to seed set across a potential gradient of pollen availability. MATERIALS AND METHODS This study was conducted during May and June 1996 in two areas located near the Queen s University Biological Station in Leeds and Grenville County, eastern Ontario, Canada. One site ( QUBS ) is a 100-ha peninsula on Lake Opinicon, 2 km west of Chaffey s Lock (44 35 N latitide, W longitude). Aquilegia canadensis is patchily distributed on the many small granite outcrops scattered throughout this area, and plants occur in a wide range of densities. The second site is located along the abandoned Lindsay Lake road ( LLR ) 8 km southwest of QUBS (44 33 N, W). Here, we used plants growing on a large 1-ha outcrop. In both populations, bumble bees and small solitary bees were the most common flower visitors, although visitors were only observed in the denser patches. Capacity for autonomous self-fertilization We used pollinator exclusion combined with hand-pollinations to quantify the capacity of A. canadensis for autonomous self-fertilization. Twenty-nine plants with at least three unopened buds were haphazardly selected (12 from QUBS and 17 from LLR) and excluded from pollinators with cylinders of coarse wire mesh covered in fine bridal veil. Each of the three buds on each plant were randomly assigned one of three pollination treatments: (1) all stigmas were hand self-pollinated at the onset of anther dehiscence; (2) stigmas were left unpollinated; or (3) stigmas were left unpollinated and all anthers were removed from the flower before they dehisced. Just after the last anther dehisced, we measured several traits for each flower receiving treatments 1 and 2, including the maximum length of a randomly chosen spur, the length of a randomly chosen sepal, style length measured as the distance from the stigma on the longest style to the point at which the nectar spurs fuse (about the basal end of the style), and the distance between the anther on the longest stamen and the stigma on the shortest style (herkogamy). Fruits were collected just before the mature follicles dehisced, and the full seeds in each were counted. The capacity for autonomous self-fertilization was assessed with a paired t test comparing seed production in flowers receiving treatments 1 and 2. Treatment 3 was included to confirm that autonomous seed set did not occur through either apomixis or accidental pollination by insects that entered the cages. Only six of the 24 emasculated flowers set fruit, and those fruits contained few seeds per carpel (X 1SE seeds, N 6) compared to flowers receiving the other two treatments ( seeds, N 48). Hence, autonomous seed production observed in A. canadensis probably occurred only via autonomous selffertilization. Timing of pollen deposition on stigmas The potential value of autonomous self-fertilization in increasing seed set was assessed by comparing the number of pollen grains on stigmas just before anther dehiscence with pollen receipt at the end of anther dehiscence. One recently opened flower was tagged on each of 46 haphazardly selected plants at QUBS. We removed two stigmas from each flower just before anther dehiscence and another two just after the last anther had dehisced. Floral

3 July 1998] ECKERT AND SCHAEFER REPRODUCTIVE ASSURANCE IN AQUILEGIA 921 measurements were performed as above. Pollen grains were counted at 100 on stigmas crushed under glass coverslips in fuschin jelly (Dafni, 1993, p. 72). Experimental manipulation of autonomous selfing The extent to which autogamous self-fertilization increases seed set was determined by removing the capacity for autogamy by emasculating flowers before anther dehiscence. We haphazardly selected 98 pairs of plants at QUBS, with the members of each pair located within 3 m of each other. The plants in each pair were matched for total number of flowers and buds (a measure of plant size) and flowering phenology. On each plant, we selected two flowers matched with two flowers on the other plant for flowering stage and sequence. For one plant in each pair, both flowers were emasculated (plant E) before anther dehiscence, whereas both flowers on the other plant were left intact (plant I). One of the two flowers on each plant was hand-pollinated with outcross pollen from a plant growing within 3 m (flower p), while the other flower was openpollinated (flower o). For most plants, the hand-cross-pollinated flower (Ep or Ip) opened earlier than the open-pollinated flower (Eo or Io). Floral measurements were performed as above, except that herkogamy could not be measured on emasculated flowers. Fruits were collected when mature and seeds counted. Thirty-seven of the 98 pairs of plants survived herbivory to provide data on seed production for all four treatments. If autogamy increases seed set, emasculated flowers should set fewer seeds than intact flowers when open-pollinated (seed set: Eo Io) but not when hand-pollinated (Ep Ip). However, two side effects of emasculation could also produce this outcome: removing anthers might reduce seed set by damaging flowers, and flowers lacking pollen may be less attractive to pollinators so that emasculated flowers set fewer seed than intact flowers because they receive fewer outcross pollen grains. With our experimental design, possible negative effects of emasculation on seed set through damage to the flowers can be estimated by comparing Ep vs. Ip flowers. Secondly, we expect that emasculation would not greatly reduce the attractiveness of flowers, since the copious nectar produced by flowers of A. canadensis appears to be the main reward for hummingbird pollinators. However, bumble bees and solitary bees were occasionally seen collecting pollen from flowers, so that emasculated flowers may have received fewer visits and less allogamous pollen than intact flowers. As it turned out, this did not pose a problem for interpreting our results. The only other drawback to this experimental design is that the level of geitonogamous (between-flower) self-pollination may be lower for Eo than Io flowers, because Ep flowers had their anthers removed, whereas Ip flowers did not. As a result, reduced seed set by Eo flowers might be caused by reduced geitonogamy in addition to the absence of autogamy. Again, this did not pose a problem in the interpretation of our results. The effects of both emasculation (E vs. I) and hand-pollination (o vs. p) as well as the interaction between these two factors were assessed using a split-plot analysis of variance (Neter, Wasserman, and Kutner, 1990, pp ) with pairs of plants as blocks, plants within pairs as plots (effect of emascuation), and flowers within plants as subplots (effect of hand pollination). Pair was a random effect and the others were fixed. The F test for Emasculation used MS Pair Emasculation as a denominator; those for all other effects used MS Residual. For this and other analyses, observed and expected mean squares were calculated using the general linear model procedure in JMP (version 3.2, SAS, 1989). The number of seeds per carpel was square-root transformed (Y Y Y 1) to meet assumptions of ANOVA. Residuals from the model did not deviate significantly from normality (Shapiro-Wilk W 0.992, N 148, P 0.89), and absolute values of the residuals were not correlated with predicted values (r 0.01, N 148, P 0.89). The extent to which autonomous self-fertilization increases seed set will depend on the availability of outcross pollen. This, in turn, can be influenced by both the level of pollinator visitation and the presence of potential mates, both of which may be affected by the density of flowering conspecifics (Lloyd, 1980). To examine whether the effect of emasculation on the seed production of open-pollinated flowers varied with plant density, we measured the number of flowering conspecifics in a 10 m radius around each experimental pair of plants (range 0 35 plants; X SD plants). We also subjectively scored mate availability on a scale of 1 (low density) to 5 (high density; X 2.8, median 3). This subjective score was strongly correlated with the density estimate (r 0.83, N 37, P ), so only the latter was used in the analyses presented below. To determine whether the effects of emasculation varied with plant density, we tested for an interaction between the effects of emasculation and plant density using an analysis of covariance performed on the seed production of openpollinated flowers only. All F tests used MS Residual as denominator. The number of seeds per carpel was square-root transformed (Y Y Y 1) to meet assumptions of ANCOVA. Residuals from the model did not deviate significantly from normality (W 0.982, N 74, P 0.71) and absolute values of the residuals were not correlated with predicted values (r 0.09, N 74, P 0.44). RESULTS Capacity for autonomous self-fertilization Flowers excluded from pollinators set almost as many seeds as flowers self-pollinated by hand, indicating a high capacity for autonomous self-fertilization. Including only plants on which at least one of the two flowers set at least one seed, unpollinated flowers produced 87.3% as many seeds per carpel (X 1SE seeds) as handpollinated flowers ( seeds; X difference seeds, paired t 0.95, df 18, one-tailed P 0.18). The relative seed production of unpollinated flowers was slightly higher (93.3% of hand-pollinated flowers) when plants on which both flowers failed to set seed were included in the analysis. The capacity for autonomous self-fertilization was associated with the proximity of stigmas and anthers during anther dehiscence. Seed production of unpollinated flowers correlated negatively with herkogamy (Fig. 1A; r 0.46, N 19, one-tailed P 0.023). Although a negative trend was also observed for hand-pollinated flowers, the correlation was weaker and not significant (r 0.32, N 18, one-tailed P 0.10). Seed production did not vary significantly with other floral traits (spur length, sepal length, and style length) for either pollinated or unpollinated flowers (all two-tailed Ps 0.08, mean P 0.50). Timing of pollen deposition on stigmas Although stigmas are presented several days before anther dehiscence, very little pollen is deposited during this time. Stigmas collected just before anther dehiscence had an average ( 1 SE) of pollen grains (range 0 19) compared to grains on stigmas at the end of anther dehiscence (range 3 615, X difference grains, paired t 7.4, df 45, one-tailed P ). Correlations between floral traits and pollen loads suggest that the relative amount of autogamous pollen deposited on stigmas increases with the proximity of stigmas and anthers. Herkogamy correlated negatively with the number of pollen grains on stigmas after anther dehiscence (Fig. 1B; r 0.24, N 46, one-tailed P 0.05) but not before anther dehiscence (r 0.06, N 46, one-tailed P 0.34).

4 922 AMERICAN JOURNAL OF BOTANY [Vol. 85 Fig. 2. The joint effects of floral emasculation and hand cross-pollination on seed production in Aquilegia canadensis. Error bars are 1 SE. Analysis of these data is in Table 1. Experimental manipulation of autonomous selfing Seed production varied considerably among flowers in the emasculation experiment (range seeds/ carpel; X 1SD seeds/carpel, N 148). Seed production varied significantly among pairs of plants (range seeds/carpel; cv 43.9%, Table 1), and was significantly higher in hand-pollinated (Ip and Ep: X 1SE seeds/carpel) than openpollinated flowers (Io and Eo: seeds/carpel, Fig. 2). However, emasculation did not affect the seed production of open-pollinated (Io Eo) or hand-pollinated flowers (Ip Ep), so the effects of emasculation and pollination did not interact (Fig. 2, Table 1). Seed production did not vary with herkogamy for intact, openpollinated flowers (r 0.03, N 37, P 0.88). Seed production of open-pollinated flowers increased significantly with plant density (Fig. 3, Table 2; r 0.39, N 74, P ). However, this effect did not differ between intact and emasculated flowers. That the positive correlation between density and seed production was not due to increased pollen availability is also supported by a positive correlation between seed production and plant density for hand-pollinated flowers (r 0.40, N 74, P ). Fig. 1. Relation between herkogamy (minimum stigma anther separation) and seed production in plants isolated from pollinators (A) and pollen loads in open-pollinated flowers (B) in Aquilegia canadensis. DISCUSSION Aquilegia canadensis has a well-developed capacity for autonomous self-fertilization that appears to result from incomplete protogyny and may be enhanced by close proximity of anthers and stigmas during pollen presentation. Accordingly, this species would appear to have what Cruden and Lyon (1989) call a facultatively xenogamous breeding system, with flowers adapted for outcrossing but equipped with a delayed selfing mechanism to ensure fertility when insufficient outcross pollen is received for full seed set. Delayed autonomous selfing would appear to be potentially valuable in terms of reproductive assurance, because stigmas receive few pollen grains before anther dehiscence. However, plants with the capacity for autonomous selfing eliminated did not set fewer seeds than intact control plants even at low densities. This suggests that autonomous selfing did not provide reproductive assurance in this population of A. canadensis during The mechanism of autonomous selfing Flowers of A. canadensis appear morphologically protogynous because they present receptive stigmas several days before anther dehiscence. However, in this population, stigmas receive very little pollen during this time (X 2.7 grains/ stigma) compared to the number of ovules available for fertilization (X 25 ovules/carpel). Therefore, protogyny TABLE 1. Split-plot analysis of variance investigating the effects of emasculation and hand cross-pollination on seed production in a natural population of Aquilegia canadensis. One plant in each matched pair was emasculated; the other was left intact. One flower on each plant was hand-outcrossed; the other was open-pollinated. For the whole model, r , F 75,72 4.9, P Treatment means are in Fig. 2. Source of variation df SS F P Pair Emasculation Pair emasculation Pollination Emasculation pollination Residual Fig. 3. The effect of the density of flowering conspecifics and floral emasculation on the seed production of open-pollinated flowers of Aquilegia canadensis. Analysis of these data is in Table 2.

5 July 1998] ECKERT AND SCHAEFER REPRODUCTIVE ASSURANCE IN AQUILEGIA 923 TABLE 2. Analysis of covariance investigating the effects of emasculation and density of flowering conspecifics on seed production of open-pollinated flowers in a population of Aquilegia canadensis. For the whole model, r , F 3,70 3.7, P Data are in Fig. 3. Source of variation df SS F P Emasculation Density Density emasculation Residual is incomplete, resulting in functional adichogamy. Consequently, there is ample opportunity for self-fertilization during anther dehiscence when pollen is shed in close proximity to the stigmas. A comparison of pollen loads among floral stages for four other eastern Ontario populations of A. canadensis revealed a similar temporal pattern of pollen deposition (K. Mavraganis and C. G. Eckert, Queen s University, unpublished data), suggesting that incomplete protogyny may be a general characteristic of A. canadensis in this part of its geographical range. Marked changes in other floral features at anther dehiscence also suggest that A. canadensis flowers are functionally adichogamous. Sepals open much farther after anther dehiscence, making the flower appear larger and exposing the contrasting yellow laminae of the nectar spurs. Also, nectar production occurs primarily after anthers have dehisced. If these changes make flowers more attractive to pollinators, it would explain how stigmas of emasculated flowers receive enough cross pollen during the 3 d of anther dehiscence to set as many seeds as intact flowers, whereas very few cross pollen grains are deposited during the 5 d before anther dehiscence. We are currently examining changes in stigma morphology and pollen tube growth during floral development to better understand the temporal pattern of pollination and fertilization in A. canadensis. In addition to dichogamy, spatial separation of stigmas and anthers (herkogamy) has been widely viewed as a floral mechanism that reduces self-pollination (Faegri and van der Pijl, 1979; Richards, 1986; Webb and Lloyd, 1986; Barrett and Eckert, 1990). Herkogamy negatively affects both autonomous seed production and the rate of self-fertilization in protandrous Aquilegia caerulea (Brunet and Eckert, 1998). Given the extensive overlap in stigma receptivity and anther dehiscence in A. canadensis, we expected that the capacity for autonomous selffertilization would be related to the proximity of anthers and stigmas. Our results support this expectation, but not unequivocally. Both seed production of flowers isolated from pollinators and the number of pollen grains on stigmas of open-pollinated flowers varied negatively with the separation between stigmas and anthers during dehiscence (Fig. 1). However, a similar though weaker correlation between herkogamy and seed production was found for hand self-pollinated flowers isolated from pollinators. At this time, the cause of this correlation is unclear. Selective value of autonomous selfing Although A. canadensis has a well-developed capacity for autonomous self-fertilization, autonomous selfing did not increase seed set in this population during the 1996 flowering season. Moreover, autonomous selfing was not particularly valuable in low density patches (Table 2, Fig. 3). These results do not support the hypothesis that autonomous selfing has been selected because it provides reproductive assurance. This is somewhat surprising considering the patchy distribution of A. canadensis at QUBS and that potential pollinators were rarely seen visiting flowers, especially in low-density patches. In addition, the substantial difference in seed production between open-pollinated and hand-pollinated flowers might suggest that pollen availability limited seed set in this population. Unfortunately, the effect of hand-pollination is confounded with potential variation in seed set among early and later-opening flowers, because the hand-pollinated flower on most plants opened earlier than the open-pollinated flower. Subsequent work in this population (C. G. Eckert, P. Thomas, and B. Massonnet, Queen s University, unpublished data) has shown a comparable decline in seed production with order of flower opening resulting from reduced ovules per carpel and seed set per ovule in later-opening flowers. This decline occurs to the same extent in both hand- and open-pollinated flowers. Moreover, plants on which all flowers were cross-pollinated did not produce more seeds compared to open-pollinated control plants. It seems unlikely, therefore, that seed production in this population is limited by the quantity of pollen deposited on stigmas. Similarly, seed production is probably not significantly influenced by pollen quality. An experiment conducted in another eastern Ontario population of A. canadensis found only minor differences in seed set between self- and cross-pollinated flowers (K. Mavraganis and C.G. Eckert, Queen s University, unpublished data). Although mechanisms for autonomous selfing in flowers that otherwise appear adapted for outcrossing have often been interpreted in terms of reproductive assurance (Cruden and Lyon, 1989), the fertility consequences of autonomous selfing have rarely been examined in natural populations. Furthermore, the few available studies indicate that autonomous selfing usually does not substantially increase seed set. The most comprehensive study to date demonstrated higher relative seed set for self-fertilizing (homostylous) vs. outcrossing (distylous) morphs in eight populations of Primula vulgaris studied over 3 yr (Piper, Charlesworth, and Charlesworth, 1986). However, a significant contrast was not observed for one population in any year of the study and was detected in only one of three years for another two populations. In springflowering Hepatica americana, floral emasculation reduced seed production by 17 50% in four North Carolina populations (Motten, 1982). However, emasculation did not reduce seed production in a New York population of the ecologically similar, spring-flowering, Hepatica acutiloba (Bernhardt, 1976). Emasculation experiments in four other species with well-developed capacities for autonomous self-fertilization all failed to demonstrate a fertility advantage associated with autonomous selfing (Calylophus serrulatus and Scilla sibrica Cruden and Lyon, 1989; Mimulus guttatus Leclerc-Potvin and Ritland, 1994; Hibiscus laevis Klips and Snow, 1997). A major problem in interpreting experimental studies

6 924 AMERICAN JOURNAL OF BOTANY [Vol. 85 of reproductive assurance is that it need not operate in all populations or during all flowering seasons for autonomous selfing to be selected. In species with ephemeral populations, autonomous selfing may be selected during colonization even though it does not increase seed set in established populations (J. Pannell and S. C. H. Barrett, University of Toronto, unpublished data). Reproductive assurance during colonization probably cannot explain the selection of autonomous selfing in A. canadensis, which typically occurs in habitats where episodes of extinction and colonization are liable to be infrequent. Autonomous selfing might also be selected if the availability of outcross pollen limits seed production in some years or populations, but not others. Accordingly, interpreting the results of experimental studies hinges on how well the spatial and temporal ranges of pollen availability have been sampled. In this study, we attempted to account for spatial variation in pollen availablity by performing emasculations in a typical population across a gradient of plant density. Ultimately, broader sampling of populations and flowering seasons will be required to confirm the results of this experiment. In the meantime, we have no evidence that the well-developed capacity for autonomous selfing benefits A. canadensis by providing reproductive assurance. More generally, our results emphasize the importance of experimental manipulations in understanding the selective value of autonomous self-pollination in flowering plants. LITERATURE CITED BAKER, H. G Self-compatibility and establishment after longdistance dispersal. Evolution 9: Support for Baker s Law as a rule. Evolution 21: BARRETT, S. C. H., AND C. ECKERT Variation and evolution of mating systems in seed plants. In S. Kawano [ed.], Biological approaches and evolutionary trends in plants, Academic Press, Tokyo., M. T. MORGAN, AND B. C. HUSBAND The dissolution of a complex genetic polymorphism: the evolution of self-fertilization in tristylous Eichhornia paniculata. Evolution 43: BERNHARDT, P The pollination ecology of Hepatica acutiloba DC. (Ranunculaceae). Bulletin of the Torrey Botanical Club 103: BRUNET, J., AND C. G. ECKERT Effect of floral morphology and display on outcrossing in blue columbine, Aquilegia caerulea (Ranunculaceae). Functional Ecology, in press. COX, P. A Baker s law: plant breeding systems and island colonization. In J. H. Bock and Y. B. Linhart [eds.], The evolutionary ecology of plants, Westview Press, Boulder, CO. CRUDEN, R. W., AND D. L. LYON Facultative xenogamy: examination of a mixed mating system. In J. H. Bock and Y. B. Linhart [eds.], The evolutionary ecology of plants, Westview Press, Boulder, CO. DAFNI, A Pollination ecology: a practical approach. IRL Press at Oxford University Press, Oxford. DARWIN, C. R The effects of cross and self-fertilization in the vegetable kingdom. John Murray, London. DOLE, J. A Role of corolla abscission in delayed self-pollination of Mimulus guttatus (Scrophulariaceae). American Journal of Botany 77: Reproductive assurance mechanisms in three taxa of the Mimulus guttatus species complex (Scrophulariaceae). American Journal of Botany 79: FAEGRI, K., AND L. VAN DER PIJL The principles of pollination ecology, 3rd ed. Pergamon Press, New York, NY. GANDERS, F. R Heterostyly, homostyly, and fecundity in Amsinkia spectabilis (Boraginaceae). Madrõno 23: GARNOCK-JONES, P. J Breeding systems and pollination in New Zealand Parahebe (Scrophulariaceae). New Zealand Journal of Botany 14: GRANT, K. A., AND V. GRANT Hummingbirds and their flowers. Columbia University Press, New York, NY. GRANT, V., AND E. J. TEMELES Foraging ability of rufous hummingbirds on hummingbird flowers and hawkmoth flowers. Proceedings of the National Academy of Sciences, USA 89: HAGERUP, O Pollination in the Faroes in spite of rain and poverty of insects. Biologiske Meddelelser Kongelige Danske Videnskabernes Selskab 18: HOLSINGER, K. E Pollination biology and the evolution of mating systems in flowering plants. Evolutionary Biology 29: INOUE, K., M. MAKI, AND M. MASUDA Evolution of Campanula flowers in relation to insect pollination on islands. In D. G. Lloyd and S. C. H. Barrett [eds.], Floral biology: studies on floral evolution in animal-pollinated plants, Chapman and Hall, New York, NY. JAIN, S. K The evolution of inbreeding in plants. Annual Reviews of Ecology and Systematics 7: KLIPS, R. A., AND A. A. SNOW Delayed autonomous self-pollination in Hibiscus laevis (Malvaceae). American Journal of Botany 84: LECLERC-POTVIN, C., AND K. RITLAND Modes of self-fertilization in Mimulus guttatus (Scrophulariaceae): a field experiment. American Journal of Botany 81: LLOYD, D. G Demographic factors and mating patterns in Angiosperms. In O. T. Solbrig [ed.], Demography and evolution in plant populations, Blackwell, Oxford, UK., AND D. J. SCHOEN Self- and cross-fertilization in plants. I. Functional dimensions. International Journal of Plant Sciences 153: MACIOR, L. W Foraging behaviour of Bombus (Hymenoptera: Apidae) in relation to Aquilegia pollination. American Journal of Botany 53: The pollination of vernal angiosperms. Oikos 30: MOTTEN, A. F Autogamy and competition for pollinators in Hepatica americana (Ranunculaceae). American Journal of Botany 69: MUNZ, P. A Aquilegia, the cultivated and wild columbines. Gentes Herbarum 7: NETER, J., W. WASSERMAN, AND M. H. KUTNER Applied linear statistical models, 3rd ed. Richard D. Irwin, Homewood, NJ. PAYSON, E. B The North American species of Aquilegia. Contributions from the National Herbarium 20: PIPER, J. G., B. CHARLESWORTH, AND D. CHARLESWORTH Breeding system evolution in Primula vulgaris and the role of reproductive assurance. Heredity 56: RATHCKE, B., AND L. REAL Autogamy and inbreeding depression in mountain laurel, Kalmia latifolia (Ericaceae). American Journal of Botany 80: RICHARDS, A. J Plant breeding systems. George Allen & Unwin, London. SAS JMP user s guide, version 2.0 edition. SAS Institute, Cary, NC. SCHEMSKE, D. W Sexual reproduction in an Illinois population of Sanguinaria canadensis L. American Midland Naturalist 100: SCHOEN, D. J., AND D. G. LLOYD Self- and cross-fertilization in plants. III. Methods for studying modes and functional aspects of self-fertilization. International Journal of Plant Sciences 153: , M. T. MORGAN, AND T. BATAILLON How does self-pollination evolve? Inferences from floral ecology and molecular genetic variation. Philosophical Transactions of the Royal Society of London, Series B 351: WEBB, C. J., AND D. G. LLOYD The avoidance of interference between the presentation of pollen and stigmas in angiosperms. II. Herkogamy. New Zealand Journal of Botany 24: WYATT, R Ecology and evolution of self-pollination in Arenaria uniflora (Caryophyllaceae). Journal of Ecology 74: