POLLINATION IN VERBASCUM THAPSUS (SCROPHULARIACEAE):

Transcription

1 American Journal of Botany 85(11): POLLINATION IN VERBASCUM THAPSUS (SCROPHULARIACEAE): THE ADVANTAGE OF BEING TALL 1 SARAH E. DONNELLY, CHRISTOPHER J. LORTIE, AND LONNIE W. AARSSEN 2 Department of Biology, Queen s University, Kingston, Ontario, Canada K7L 3N6 According to the effective pollination hypothesis, tall stature resulting from strong apical dominance attracts greater pollinator visitation, thus allowing larger pollen loads and/or greater outcrossing rates, which in turn produces more vigorous offspring with greater genotypic variability and/or less inbreeding depression. Components of this hypothesis were tested in Verbascum thapsus, which commonly grows unbranched to over 2 m tall with strong apical dominance suppressing all axillary meristems. A natural population survey indicated that plants with visiting pollinators were significantly taller than their nearest neighboring individuals not possessing a visiting pollinator. Plants in natural populations with excluded pollinators produced seeds via a delayed selfing mechanism. However, delayed selfing under pollinator exclusion resulted in only 75% of the seed set obtained with natural pollinators. Under natural pollination, emasculated flowers experienced a 50% reduction in pollen deposition by the time of flower closure but only a 5% reduction in seed set relative to intact flowers. Hence, taller plants attracted more pollinators and maximum seed set could not be achieved without pollinators. Comparison of seed set and seed mass in plants that were artificially selfed and artificially crossed (in both the greenhouse and in natural populations) indicated that plants were fully self-compatible with no evidence of early-acting inbreeding depression. However, this does not exclude the possibility that inbreeding depression is manifested in later life stages. The results suggest that V. thapsus has a mixed mating system with potential for reproductive assurance and various levels of outcrossing depending on variables affecting pollinator availability (e.g., population size). Key words: height; inbreeding depression; outcrossing; Scrophulariaceae; seed mass, seed set; self-compatibility; Verbascum thapsus. Several studies have investigated the role of pollinator attractants such as flower color (e.g., Delph and Lively, 1989), flower size (e.g., Johnson, Delph, and Elderkin, 1995) floral display size (e.g., Robertson and Macnair, 1995) and morphological symmetry (e.g., Moller and Eriksson, 1995). Few studies, however, have investigated whether plant height affects attractiveness to pollinators (Schaffer and Schaffer, 1977; Larson and Larson, 1990; Peakall and Handel, 1993). The height of a plant is determined primarily by internode length and the strength of apical dominance, i.e., the suppression of axillary meristems by the apical meristem, thus promoting vertical extension in upright plants at the expense of lateral growth (Cline, 1991). The effective pollination hypothesis for apical dominance predicts that individuals benefit from strong apical dominance because pollinators are more attracted to taller plants, which thus receive greater pollen deposition and/or greater outcrossing rates resulting in progeny with greater genotypic variability and/or reduced inbreeding depression (Aarssen, 1995). There are two main assumptions of the effective pollination hypothesis: (1) competition for pollinator service exists, and (2) increased pollinator service provides fitness benefits in terms of greater pollen quantity and/or 1 Manuscript received 18 July 1997; revision accepted 23 April Field assistance was provided by Steven Bonser, Erin Corness, Christopher McPhee, and Sally Taylor. Helpful comments at various stages of this research were provided by Christopher Eckert. This research was supported by the Natural Sciences and Engineering Research Council of Canada through a research grant to LWA. 2 Author for correspondence higher pollen quality (increased outcrossing and reduced selfing). The effect of pollinator service on selfing rate may depend on the mode of selfing that the plant possesses (Lloyd, 1979). Prior self-fertilization involves selfing before a flower opens and before stigmas are exposed for receipt of outcross pollen. Competing selffertilization involves competition between self-pollen and outcross pollen. Finally, in delayed self-fertilization, a flower adds self-pollen to the existing pollen load once potential for outcrossing has ceased. Delayed selfing is always advantageous in monocarpic species since it does not interfere with crossing and acts as a better than nothing alternative even when inbreeding depression is high. In the present study, we explored components of the effective pollination hypothesis for apical dominance using Verbascum thapsus L. (common mullein). This species is among the tallest of monocarpic herbaceous species in North America, commonly growing to heights of 2 m or more from a single stem. It rarely branches despite numerous axillary meristems associated with shortly spaced leaves along the stem. Hence, its tall, sparsely branched architecture is a product of very strong apical dominance (Lortie and Aarssen, 1997). The main stalk and any branches that are produced terminate in a single indeterminate inflorescence spike of bright yellow flowers that are visited by pollinators, primarily bees. In a field survey, we analyzed the relationship between plant height and pollinator visitation. Pollen deposition, seed set, and seed mass of emasculated vs. intact flowers were compared to assess the role of pollinators and the im-

2 November 1998] DONNELLY ET AL. APICAL DOMINANCE AND POLLINATION 1619 portance of prior and delayed selfing in V. thapsus. Finally, we recorded seed mass and seed set derived from artificial self- and artificial cross-pollinations to evaluate self compatibility and to test for early-acting inbreeding depression. MATERIALS AND METHODS Study species Verbascum thapsus is a Eurasian species, introduced in the 1700s and now distributed throughout North America with northern limits in southern Canada (Reinartz, 1984a). It is an early colonizer of disturbed habitats, requiring bare soils and high light exposure for spring germination (Semenza, Young, and Evans, 1978; Gross, 1980). It is normally a biennial (but sometimes an annual or triennial), overwintering as a rosette and bolting in its second growing season with flowers blooming in late June to early August (Gross and Werner, 1978; Reinartz, 1984a). Hence, the species usually establishes under relatively uncrowded conditions where its tall stature is not readily explained as a product of selection from competition for light. Because Verbascum is strictly monocarpic, however, the mature life history stage is ephemeral, commonly undergoing local exclusion after a single generation due to invading species (Gross, 1980). Allocation to reproductive biomass is small in V. thapsus relative to other monocarpic species (Reinartz, 1984b). A high investment in structural biomass may represent an adaptation to its commonly impoverished habitat conditions, or the tall stalk of V. thapsus may serve to attract pollinators and aid in seed dispersal (Reinartz, 1984b). Flowers mature in a spiral from the bottom of the inflorescence to the top. Each individual flower lasts only until midday, at which time a delayed selfing mechanism is assumed to be implemented when corollas dehisce (Gross and Werner, 1978). Autumn seed dispersal is limited spatially; 75% of the seeds land within 1 m of the parent plant (Gross and Werner, 1978). However, fruits retain many seeds over the winter when dispersal may occur more extensively over ice and snow. Study site Quarries and disturbed fields were surveyed for prospective study populations of V. thapsus in the vicinity of Kingston, Ontario during May Two sites were selected in Kingston Township (population A N,76 35 W; population B N, W) based on their proximity to one another ( 30 km) and large population size ( 500 plants). Verbascum thapsus was the dominant species at both sites. Since it was essential that flower manipulation occurred before arrival of pollinators (see below), both sites had to be treated early in the morning and hence needed to be located nearby. As well, a large population size was desirable to ensure that experimentation on selected plants would not disrupt pollination dynamics of the population as a whole. In a small population, manipulation of all the plants might severely interfere with pollinator activity and natural pollen flow. A large population provided many untreated buffer plants. Both sites were spot sprayed twice with pesticide (Dursban 4E) to eliminate weevil seed predation. Spraying was conducted in early June before flowering and associated pollinator activity was prevalent. Some weevils were sighted throughout the growing season even after two applications of pesticide and were removed from the plant by hand when noticed. Survey of plant heights in relation to pollinator visitation A survey of site A was performed on 20 July 1995 and 3 August 1995 between 0730 and While systematically walking through the site, we observed individuals being visited by pollinators. When a pollinator was sighted on a given (target) individual, its nearest neighbor that did not have a pollinator at the same time was identified. Plant height, inflorescence length, number of branches, and number of open flowers were recorded on both the target plant and the neighbor. The distance between the target and neighbor was also recorded. A total of 50 target/ neighbor pairs were recorded on each survey day. Care was taken not to sample the same target/neighbor pair more than once. Observer bias was minimal since sampling was not done from a distance but from within the patch. Hence, height differences could not be readily discerned when identifying pairs. Inflorescences of most plants were near or above eye level, reducing the possibility of pollinators on taller plants being observed preferentially. As well, since flowers open in spiral arrays up the spike, pollinator foraging position on the inflorescence is largely independent of plant height, i.e., a pollinator on a tall plant could be foraging below the height of a neighboring shorter plant and hence would not be necessarily easier to detect. Survey of pollen deposition Five individuals were selected randomly and tagged at site A. On each treatment day, a pair of flowers per plant was chosen based on their proximity. This was done to reduce the effect of inflorescence position, which may bias pollinator visitation. One flower was emasculated (just after flower opening) and the other was left intact. Treatments were applied for an average of 8 d per plant, throughout the flowering period (13 July to 8 August 1995). When a plant did not have any flowers open on a given treatment day, treatment was prevented and hence the total number of treatment days for such an individual was reduced. The number of open flowers was counted on each day to test for the effect of this variable on pollen deposition. Stigmas were harvested once pollinator activity had ended for the day, but before the flowers had closed and hence before delayed selfing occurred. Slides were prepared in the field. A small cube of basic fuschin jelly (Kearns and Inouye, 1993) was placed on a clean slide and a fresh stigma then placed onto the surface of the jelly. A flame was held under the slide to melt the fuschin jelly, a cover slip was placed over the jelly and the stigma was crushed with firm pressure onto the cover slip. Crushing the stigma to a single-cell layer enabled all pollen grains to be distinguished and counted. Pollen grains were counted under a microscope by systematically scanning the complete surface area of the jelly. These treatments parallel those of treatments four (E O) and five (O) of the pollination experiments (see below) permitting a test of the effects of treatment on both pollen deposition and subsequent seed set. One stigma was harvested from each of ten haphazardly chosen individuals on 14 August 1995, just after flowers had opened. Stigmas were selected from the flowers that appeared (by visual inspection) to not yet have pollen deposited onto the stigmatic surface (i.e., clean ). This was done to test the assumption that stigmas possessed little or no pollen prior to the application of treatments (i.e., no prior selfing). Stigmas were harvested haphazardly from those that appeared to be clean. Stigmas were prepared and pollen grains were counted as described above. Pollination experiments Pollination experiments were conducted at both field sites and in a greenhouse. In the field, 30 individuals were randomly selected and tagged at each site; five plants were grown from seed for greenhouse use. In the greenhouse, plants were watered daily, nutrients were added intermittently, and plants were placed in full sunlight. Five treatments were designed to control pollen transfer so that the components of pollen sources were regulated. Two treatments involved artificial pollination, whereby either outcross (X) or self (S) pollen was introduced to the stigma of the target flower. Subsequently, anthers were removed (emasculation) (E) immediately (i.e., early morning) to prevent delayed selfing when corollas dehisced. These techniques resulted in artificial self (E S) and artificial cross (E X) treatments. The third treatment involved leaving the flower intact, permitting delayed or natural selfing (S) only. These first three treatments were conducted in the absence of pollinator visitation and hence were easily manipulated within the greenhouse. In the field, pollinator exclusion bags were used (made from fine cloth mesh). Wire frames were designed to prevent the bags from brushing against the flowers and interfering with pollination treatments. All flowers open on a given day in the greenhouse were treated and

3 1620 AMERICAN JOURNAL OF BOTANY [Vol. 85 Fig. 1. Scatterplot of heights in V. thapsus for pollinator-visited plants ( targets ) vs. their nearest unvisited neighboring plants (paired t test; t 4.36, P , df 99). Both variables were normally distributed. Data below the leading diagonal line represent target plants that were taller than their unvisited neighbors. Fig. 2. Comparison of mean flower number ( SE) in V. thapsus for pollinator-visited ( target ) plants vs. their nearest unvisited neighboring plants (Wilcoxon signed ranks; W , P , df 1). divided equally among the treatments. This resulted in a total of flowers per treatment per plant from 11 May to 9 June Each plant produced 90 flowers over the flowering season. The fourth and fifth treatments involved full exposure to natural pollinators (open) (O) and hence were conducted in the field sites only. Treatment 4 involved emasculation (E O) and treatment 5 involved leaving the flower intact (O). Flowers were chosen and assigned to treatments haphazardly from open flowers that did not yet appear to have any pollen deposited onto the stigma (i.e., clean stigmas). Each individual in the field was subjected to all five treatments over the peak flowering season (5 July 3 August 1995). If an individual had no open flowers on a given treatment day, treatment was prevented for that plant. The treatments were rotated over 3-d cycles, spanning 16 d for site A and 13 d for site B. Hence, the 30 plants were divided into three groups of ten and a maximum of 25 flowers were treated per plant (i.e., each plant had five flowers affected by each of the five treatments). For example, on day 1, a plant would be bagged and undergo treatments (E S), (E X) and (S). On day 2, the bag would be removed and the plant would have a single flower emasculated (E O). Finally, on day 3, the bag would remain off the plant and the flowers would be left intact (O). The calyx of each flower treated was marked with nontoxic, acrylic paint with each color corresponding to treatment. To prevent pollinator interference, treatments were applied early in the morning, just as flowers were opening (first site, 0515; second site, 0630). The order of site visitation was alternated to reduce the possible effects of delayed treatment. Such a delay could permit pollinator activity before treatments were implemented and hence interfere with regulated pollen deposition. However, once pollinator activity had ceased for the day, bags were transferred to the appropriate plants at the site that would be visited second on the following day. As a result, only the emasculated (E O) treatment could be affected, in which case pollinator activity may be slightly underestimated (i.e., bags were still on the plants when pollinator activity had commenced). Flowers closed at 1100, depending on weather conditions. Hence, it was assumed that daily pollinator activity was long enough for the influence of this lag time to be relatively small. In both experiments, fruits were harvested at maturity, just prior to dehiscence and number of seeds per fruit were counted. For plants in the field experiment, weevil damage (if any) was indexed by recording the number of weevils per fruit (two, one, or none) and the life stage of the weevil (adult, juvenile, or pupae). Due to weevil damage, replication of treatments per plant was reduced to zero or one in some cases. Hence, mean seed number per treatment for each plant was used in the analyses. Plant was not analyzed as a main effect due to this reduced replication. Fifteen of the 30 plants in each site were haphazardly chosen to compare seed mass from undamaged fruits in the artificial self vs. artificial cross treatments. Twenty seeds were selected randomly from each fruit and weighed. Mean seed mass was calculated by dividing this value by 20. TABLE 1. Spearman rank correlation analyses (r s ) of traits in V. thapsus for (A) plants observed to be visited by a pollinator ( target plants) and (B) their nearest unvisited neighboring individuals. Nonparametric analysis was used in each case because one or both variables was not normally distributed. (df 99 and P in all cases). Characteristic RESULTS Survey of plant heights in relation to pollinator visitation Plants visited by pollinators (targets) were significantly taller (mean SE cm) than their nearest unvisited neighbors (mean SE cm) (Fig. 1). Target plants also had significantly more branches (Wilcoxon signed ranks; W 602.0, P , df 1), longer inflorescences (Wilcoxon signed rank; W , P , df 1) and more open flowers (Wilcoxon signed ranks; W , P , df 1) (Fig. 2). Plant height, number of branches, inflorescence length, and number of Plant height Inflorescence length Number of open flowers Number of branches A) Target plants Plant height Inflorescence length Number of open flowers Number of branches B) Neighbor plants Plant height Inflorescence length Number of open flowers Number of branches

4 November 1998] DONNELLY ET AL. APICAL DOMINANCE AND POLLINATION 1621 Fig. 3. Mean pollen quantity ( SE) on stigmas of V. thapsus harvested before delayed selfing in the field under two treatments [emasculated with natural pollination (E O) and intact with natural pollination (O)], and on clean stigmas immediately after flower opening (prior selfing) (Mann-Whitney rank sum; U 34.88, P , df 2). Means with different letter codes are significantly different (P 0.05) based on a Tukey-Kramer post hoc test. open flowers were all positively correlated with each other for both the target plants and the neighbors (Table 1). Survey of pollen deposition The number of flowers open on a given day did not affect pollen deposition for either intact (linear regression; r , P , df 35) or emasculated treatments (linear regression, r , P , df 35) and hence was not included as a variable in the analyses. Mean pollen load deposited on stigmas of intact (O) flowers was significantly greater than those that were emasculated (E O) (paired t test; t 13.43, P , df 4). Emasculated flowers suffered a 50% (based on median values) reduction in pollen deposition compared to intact flowers (Fig. 3). Pollen load on clean stigmas (i.e., appearing to lack pollen) from prior selfing was small (mean SE ) was significantly less than on stigmas in the (O) or (E O) treatments (Fig. 3). Fig. 4. Mean seed set per fruit ( SE) for V. thapsus under three pollination treatments in the greenhouse: artificial self (E S); artificial cross (E X); and natural (delayed) self (S). Means with the same letter code are not significantly different (P 0.05) based on a Tukey- Kramer post hoc test. differences existed between treatments (Table 2). The natural/delayed self (S) treatment yielded significantly fewer seeds per fruit than both artificial self (E S) (69.3 %) and artificial cross (E X)(64.4%) treatments (Table 2, Fig. 4). There was no difference in seed set between the (E S) (mean SE ) and (E X) (mean SE ) treatments. However, the effect of treatment varied among plants (plant treatment interaction) (Table 2, Fig. 5). Natural populations The artificial self (E S) treatment did not differ significantly in number of seeds per fruit from the artificial cross (E X) treatment (repeatedmeasures [RM] ANOVA; F , P , df 1) (Fig. 6a) or the open (O) treatment (RM ANOVA; F , P , df 1) (Fig. 6b). However, the Pollination experiments Greenhouse Significant TABLE 2. Analysis of factors affecting seed set in greenhouse plants of V. thapsus using a mixed-model two-way ANOVA. Main effects were plant (random) and treatment (fixed) with plant by treatment interaction. Pollination treatments were artificial self (E S), artificial cross (E X), and natural/delayed self (S). Model fit: r Effect df SS MS F P Plant Treatment Plant Treatment Model Error Total Fig. 5. Mean seed set per fruit ( SE) for each of five greenhouse plants of V. thapsus, under each of three treatments [artificial self (E S); artificial cross (E X); natural (delayed) self (S)] (repeated-measures ANOVA; F 2.600; P , df 8).

5 1622 AMERICAN JOURNAL OF BOTANY [Vol. 85 Fig. 6. Comparison of mean seed set per fruit ( SE) between various treatment pairs in V. thapsus from two natural populations [artificial selfing (E S); artificial crossing (E X); natural (delayed) selfing (S); emasculated with natural pollination (E O); intact with natural pollination (O)]. Data from the two populations were pooled. A significant difference (P 0.05) between treatments is indicated by an asterisk based on repeated-measures ANOVA. artificial self treatment yielded significantly more seeds than the natural/delayed self treatment (S) (RM ANOVA; F 18.76, P , df 1) (Fig. 6c). The open treatment also produced significantly more seeds than the natural selfing treatment (RM ANOVA; F , P , df 2) (Fig. 6d). The open treatment produced significantly more seeds than the emasculated treatment (E O) (RM ANOVA; F 9.438, P , df 1) (Fig. 6e). Finally, the emasculated treatment (E O) produced significantly more seeds per fruit than the delayed selfing treatment (S) (RM ANOVA; F , P , df 1) (Fig. 6f). All of the analyses revealed a significant plant (within population) effect (Table 3). Plants in population A consistently produced more seeds per fruits than those in population B, although the population effect was not significant in all comparisons (Table 3). There was no difference in seed mass between artificial self (E S) (mean SE g) and artificial cross (mean SE g) (E X) treatments (Table 4). However, variation in seed mass between plants (within population) was highly significant (Table 4). DISCUSSION Pollinator visitation and plant height Plants of V. thapsus that had visiting pollinators were generally taller than their nearest unvisited neighbors (Fig. 1). Taller plants, however, had more open flowers (Fig. 2) owing to more branches and longer inflorescence spikes (Table 1). The relative attractiveness of stalk height and number of open flowers is difficult to distinguish (Hainsworth, Wolf, and Mercier, 1984). It is unclear whether pollinators are generally attracted to a more conspicuous (taller) plant or whether pollinators associate a taller plant with increased reward under optimal foraging theory (Schaffer and Schaffer, 1977, 1979; Larson and Larson, 1990). An increase in pollinator visitation frequency to individuals has been generally assumed to increase the proportion of flowers visited and this has been reported in some cases (e.g., Willson and Price, 1977). However, some research indicates that the proportion of flowers visited actually decreases with inflorescence size, resulting in per flower visitation rates equal to plants with fewer flowers (Schmidt-Hempel and Speiser, 1988; Rodriguez-Robles, Melendez, and Ackerman, 1992; Robertson and Macnair, 1995). This is predicted in the ideal free distribution foraging model, where pollinators forage systematically to equalize gains per flower (Dreisig, 1995). Hence, plants with more flowers may experience a fecundity gain due to a higher fruit production but may not receive increased pollinator visitation rates per flower. Nevertheless, since taller individuals do receive greater pollinator visitation rates per plant, the resulting seed crop may have greater genotypic diversity. Further research is required to test this prediction. The interpretation of greater branching in taller plants requires further study. Branching may be energetically affordable in only larger plants or branching may be a compensatory response to fruit infestation by weevils (Lortie and Aarssen, 1997). Branching may also be size dependent in the adaptive sense that fitness may be maximized by investing in only vertical extension when small (to maximize the chances of attaining a dominant vertical position within the population) and adding lateral growth through branching (to maximize floral display and seed production) only after the plant has already reached a tall stature. Pollen quantity and seed set Flowers that were artificially pollinated did not produce more seeds than intact flowers with natural pollination (Fig. 6a, b). However, intact naturally pollinated flowers produced significantly more seeds than flowers that were allowed only natural (delayed) selfing (Fig. 6d). Flowers with artificial selfing also produced more seeds than flowers with natural (delayed) selfing (Fig. 6c). Hence, under natural pollination, plants were not pollen limited. However, flowers with only delayed selfing (in the absence of pollinators) were pollen limited producing only 75% of maximum seed set; full seed set was not realized without pollinators. Under natural pollination, emasculated flowers received a 50% reduction in pollen deposition by the time of flower closure (Fig. 3) but only a 5% reduction in seed set (Fig. 6e) relative to intact flowers. Hence, natural levels of autogamous pollen comprised 50% of the total pollen load even before delayed selfing. Pollinators, however, may be less attracted to emasculated flowers, particularly when primarily harvesting pollen (Young and

6 November 1998] DONNELLY ET AL. APICAL DOMINANCE AND POLLINATION 1623 TABLE 3. Analysis of factors affecting mean seed set in natural populations of V. thapsus using repeated-measures ANOVA. Main effects are population (fixed) and treatment (fixed) with plant nested in population effect (hierarchical). Interaction was limited to treatment by population. Replication of treatments within plants was too low to include plant as a main effect (random) or interaction effect. The following treatment pairs were compared (see text for definition): (A) E S vs. E X; (B) E S vs. O; (C) E S vs. S; (D) S vs. O; (E) E O vs. O; (F) E O vs. S. Effect df SS MS F P A) E S vs. E X r Population Treatment Plant [Population] Treatment Population Model Error Total B) E S vs. O r Population Treatment Plant [Population] Treatment Population Model Error Total C) E S vs. S r Population Treatment Plant [Population] Treatment Population Model Error Total D) O vs. S r Population Treatment Plant [Population] Treatment Population Model Error Total E) O vs. E O r Population Treatment Plant [Population] Treatment Population Model Error Total F) O E vs. S r Population Treatment Plant [Population] Treatment Population Model Error Total Stanton, 1990). This effect may be significant given the low nectar levels in V. thapsus and that pollinators (honey bees and bumble bees) appear to forage for only pollen (S. Donnelly, personal observation). If so, the level of autogamous pollen may have been overestimated and the role of pollinators underestimated. Pollen quality Self pollen is of lower quality to a plant than outcross pollen when inbreeding depression ensues. Although inbreeding depression in V. thapsus was not detected in this study through seed set or seed mass, this conclusion has limitations. Clearly, it is important to sample indices of fitness throughout the life cycle of progeny. Inbreeding depression may be detectable only in later life stages (e.g., Belaoussoff and Shore, 1995). Inbreeding depression also may not be detectable unless the plant undergoes stress (Eckert and Barrett, 1994). Environmental conditions in the field sites were not measured, although each population was large, of high density and possessed large individuals; relatively

7 1624 AMERICAN JOURNAL OF BOTANY [Vol. 85 TABLE 4. Analysis of factors affecting mean seed mass in natural populations of V. thapsus using repeated-measures ANOVA. Main effects are population (fixed) and treatment [emasculated and selfed (E S) vs. emasculated and crossed (E X)] (fixed) with plant nested in population effect (hierarchical). Interaction was limited to treatment by population. Replication of treatments within plants was too low to include plant as a main effect (random) or interaction effect. Model fit: r Effect df SS MS F P Population Treatment Plant [Population] Treatment Population Model Error Total low stress may be generally implied from their success. Conditions within the greenhouse were also benign. Darwin (1876) detected inbreeding depression in V. thapsus, using height as an estimate of fitness after several generations of selfing. This is of particular interest, under the effective pollination hypothesis for apical dominance (Aarssen, 1995). If tall plants have a fitness advantage via artificial outcrossing, it follows that tall plants that occur naturally may also have been progeny of highly outcrossed parentals. Under the effective pollination hypothesis, the parents may have been tall to gain these outcross pollinations through increased pollinator service. Here, height does not necessarily have to be a heritable trait per se, but simply an index of fitness reflecting levels of inbreeding. When cross pollinations were performed, anthers were used (for pollen transfer) from individuals within a convenient distance (1 2 m). Hence, we did not account for biparental inbreeding associated with crossing closely spaced individuals; this can produce a conservative estimate of inbreeding depression. However, since seed dormancy enables seed dispersal in time, all of the individuals of a given cohort are unlikely to have been produced from the same parental generation. Therefore, distances between individuals do not necessarily reflect degree of relatedness. Population size and density The populations selected for this study were atypical as V. thapsus populations are often small and isolated (Gross, 1980). Population size and density probably affect pollinator attraction and degree of plant competition for pollinator service (Sih and Baltus, 1987). The cost of within-plant pollinator movement is low in dense populations (Roberston and Macnair, 1995). As a result, pollinators are able to visit many plants instead of just the most attractive and hence the benefit of attractive features is reduced. This permits an increased number of visitations per plant and the potential to maintain equalized visitation to all flowers, under the ideal free distribution hypothesis (Dreisig, 1995). In this study, pollen deposition was not influenced by flower number, which may be attributed to high population density. Decreased outcrossing rates may result from decreases in a population size and density, inferred through seed set (Kunin, 1993) and electrophoretic analysis (Farris and Mitton, 1984; Treuren et al., 1993). These studies support the theoretical modeling of Inguarsson and Lundberg (1995), which proposes that pollinator service is not limiting in high-density environments. Pollinator limitation would, however, arise when pollinators are rare. Such circumstances arise when a population is too small to attract pollinators. Hence, the delayed selfing mechanism may be crucial for V. thapsus when growing in small, isolated populations where pollinator service may be limited and reproductive assurance is required. The proportion of autogamous pollen on stigmas in such populations may be expected to be greater than reported in the present study. Male vs. female reproductive success Pollinator service may benefit male as well as female reproductive success. In this study, only female RS was estimated via seed set and pollen deposition. However, increased male reproductive success is often positively correlated with attractive floral traits, measured through various indices (e.g., pollen removal). Often only an improvement in male reproductive success is realized (Willson and Price, 1977; Young and Stanton, 1990). Further study is required to test whether the benefits of tall stature under the effective pollination hypothesis for apical dominance are realized through greater male reproductive success. LITERATURE CITED AARSSEN, L. W Hypotheses for the evolution of apical dominance in plants: implications for the interpretation of overcompensation. Oikos 74: BELAOUSSOFF, S., AND J. S. SHORE Floral correlates and fitness consequences of mating-system variation in Turnera ulmifolia. Evolution 49: CLINE, M. G Apical dominance.botanical Review 57: DARWIN, C The effects of cross and self fertilization in the vegetable kingdom. John Murray, London. DELPH, L. F., AND C. M. LIVELY The evolution of floral colour change: pollinator attraction versus physiological constraints in Fuchia excorticata. Evolution 43: DREISIG, H Ideal free distributions of nectar foraging bumblebees. Oikos 72: ECKERT, C. G., AND S. C. H. BARRETT Inbreeding depression in partially self-fertilizing Decadon verticillatus (Lythraceae): population-genetic and experimental analyses. Evolution 48: FARRIS, M. A., AND J. B. MITTON Population density, outcrossing rate, and heterozygote superiority in Ponderosa Pine. Evolution 38: GROSS, K. L Colonization by Verbascum thapsus of an old-field in Michigan. Journal of Ecology 68: , AND P. A. WERNER The biology of Canadian weeds. 28. Verbascum thapsus L. and V. blatteria L. Canadian Journal of Plant Science 58: HAINSWORTH, F. R., L. L. WOLF, AND T. MERCIER Pollination and pre-dispersal seed predation: net effects of reproduction and inflorescence characteristics in Ipomopsis aggregata. Oecologia 63: INGUARSSON, P. K., AND S. LUNDBERG Pollinator functional response and plant population dynamics: Pollinator as a limiting resource. Evolutionary Ecology 9: JOHNSON, G. S., L. F. DELPH, AND C. L. ELDERKIN The effect of petal-size manipulation on pollen removal, seed set, and insectvisitor behaviour in Campanula ameriacana. Oecologia 102: KEARNS, C. A., AND D. W. INOUYE Techniques for pollination biologists. University Press of Colorado, Niwot, CO.

8 November 1998] DONNELLY ET AL. APICAL DOMINANCE AND POLLINATION 1625 KUNIN, W Sex and the single mustard: population density and pollinator behaviour effects on seed-set. Ecology 74: LARSON, K. S., AND R. J. LARSON Lure of the locks: showiest ladies-tresses orchids, Siranthes romanzoffiana, affect bumblebee, Bombus spp., foraging behaviour. Canadian Field-Naturalist 104: LLOYD, D. G Some reproductive factors affecting the selection of self-fertilization in plants. American Naturalist 113: LORTIE C. J., AND L. W. AARSSEN Apical dominance as an adaptation in Verbascum thapsus: effects of water and nutrients on branching. International Journal of Plant Sciences 158: MOLLER, A. P., AND M. ERIKSSON Pollinator preference for symmetrical flowers and sexual selection in plants. Oikos 73: PEAKALL, R., AND S. N. HANDEL Pollinators discriminate among floral heights of a sexually deceptive orchid: implications for selection. Evolution 47: REINARTZ, J. A. 1984a. Life history variation of common mullein (Verbascum thapsus) I. Latitudinal differences in population dynamics and timing of reproduction. Journal of Ecology 72: b. Life history variation of common mullein (Verbascum thapsus). II. Plant size, biomass partitioning and morphology. Journal of Ecology 72: ROBERTSON, A. W., AND M. R. MACNAIR The effects of floral display size on pollinator service to individual flowers of Myosotis and Mimulus. Oikos 72: RODRIGUEZ-ROBLES, J. A., E. J. MELENDEZ, AND J. D. ACKERMAN Effects of display size, flowering phenology, and nectar availability on effective visitation frequency in Comparettia falcata (Orchidaceae). American Journal of Botany 79: SCHAFFER, W., AND M. V. SCHAFFER The adaptive significance of variations in reproductive habit in the Agavaceae. In B. Stonehouse and C. Perrins [eds.], Evolutionary ecology, University Park Press, London., AND The adaptive significance in reproductive habit in the Agavaceae. II. Pollinator foraging behaviour and selection for increasedreproductive expenditure. Ecology 69: SCHMID-HEMPEL, P., AND B. SPEISER Effects of inflorescence size on pollination in Epilobium angustifolium. Oikos 53: SEMENZA, R. J., J. A. YOUNG, AND R. A. EVANS Influence of light and temperature on the germination and seedbed ecology of common mullein (Verbascum thapsus). Weed Science 26: SIH, A., AND M. BALTUS Patch size, pollinator behavior, and pollinator limitation in catnip. Ecology 68: TREUREN, R. V., R. BULSMA, N. J. OUBORG, AND W. V. DELDEN The effects of population size and plant density on outcrossing rates in locally endangered Salvia pratensis. Evolution 47: WILLSON, M. F., AND P. W. PRICE The evolution of inflorescence size in Asclepias (Asclepiadaceae). Evolution 31: YOUNG, H. J., AND M. L. STANTON Influences of floral variation on pollen removal and seed production in wild radish. Ecology 71: