D1STYLY, P O L L E N FLOW AND SEED SET IN MENYANTHES TR1FOLIATA ( M E N Y A N T H A C E A E ) by N A N C Y L Y N N E CHRISTY B .Sc , The University of British Columbia, 1983 A THESIS SUBMITTED IN P A R T I A L F U L F I L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF M A S T E R OF SCIENCE in T H E F A C U L T Y OF G R A D U A T E STUDIES Department of Botany We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH C O L U M B I A March 1987 ® Nancy Lynne Christy, 1987 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at The University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his or her representatives. It is understood that, copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Botany The University of British Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date: March 1987 A B S T R A C T The influence of variation in style length on pollen flow and seed set was examined in six populations of Menyanthes trifoliata in southwestern British Columbia to evaluate Ganders' hypothesis that morphological distyly increases the fecundity of a diallelic self-incompatible plant. In five populations, Menyanthes was distylous and self-incompatible. The sixth population consisted of pins and morphological homostyles (thrums with unusually long styles). In each population morph frequency, pollen frequency, the composition of stigmatic pollen loads and seed set were estimated. Results from the six populations demonstrate that the size and composition of stigmatic pollen loads fluctuates erratically during the flowering season. Pins and thrums experienced disassortative pollination, assortative pollination and random pollination at different times during the season. Homostyles were always assortatively pollinated. Among populations there was a high correlation between morph frequency (pollen frequency) and the composition of stigmatic pollen loads. However, in anisoplethic populations of Menyanthes morphological distyly seems to compensate for the rarity of one floral form, by increasing the proportion of compatible pollen received by the opposite floral form. Comparison of the composition of stigmatic pollen loads of homostyles with those of thrums revealed that the separation of stigmas and anthers in distylous flowers of Menyanthes reduces the number of incompatible pollen grains received, and the reciprocal placement of stigmas and anthers appears to increase the number of compatible pollen grains received. Seed set in the six populations of Menyanthes was always below the potential maximum. Among populations there was a high correlation between the number of compatible pollen grains received and seed set. In at least one population, pollen availability was a major factor ii limiting seed set, but other factors are probably influencing seed set in Menyanthes as well. Thrums set significantly more seeds per capsule than homostyles. In populations of Menyanthes, the reciprocal placement of stigmas and anthers in distylous flowers (compared to homostylous flowers) increases the amount of compatible pollen deposited on stigmas, and this increase is associated with greater fecundity. iii TABLE OF CONTENTS ABSTRACT ii LIST OF TABLES v LIST OF FIGURES vi ACKNOWLEDGEMENT viii I. INTRODUCTION 1 II. FLORAL MORPHOLOGY ! 6 A. INTRODUCTION 6 B. MATERIALS AND METHODS 10 C. RESULTS 13 D. DISCUSSION 22 III. EXPERIMENTAL POLLINATIONS 28 A. INTRODUCTION 28 B. MATERIALS AND METHODS 31 C. RESULTS 33 D. DISCUSSION 37 IV. POLLEN FLOW 42 A. INTRODUCTION 42 B. MATERIALS AND METHODS 45 C. RESULTS 53 D. DISCUSSION ; 76 V. SEED SET 82 A. INTRODUCTION 82 B. MATERIALS AND METHODS 84 C. RESULTS 87 D. DISCUSSION 101 VI. CONCLUSIONS 107 BIBLIOGRAPHY 114 iv L I S T O F T A B L E S 1. Floral characters of pins, thrums and homostyles of Menyanthes trifoliata. ... 14 2. Variance components of floral characters of Menyanthes trifoliata 15 3. Number and percentage of pollen grains removed from pin and thrum flowers of Menyanthes trifoliata by pollinators 20 4. Seed set following self, geitonogamous, intramorph and intermorph pollinations of pins and thrums of Menyanthes trifoliata 34 5. Representation of floral morphs in six populations of Menyanthes trifoliata. ... 54 6. Frequency of pin, thrum and homostyle pollen produced in six populations of Menyanthes trifoliata 56 7. Percentage of pin and thrum (or homostyle) pollen participating in compatible pollinations in six populations of Menyanthes trifoliata 57 8. Percentage of pollen on stigmas of pin, thrum and homostyle flowers of Menyanthes trifoliata from six populations 58 9. Composition of stigmatic pollen loads of pin, thrum and homostyle flowers of Menyanthes trifoliata from six populations 63 10. Expected and observed pollen frequencies on stigmas of pin and thrum flowers of Menyanthes trifoliata from population BL2 71 11. Expected and observed pollen frequencies on stigmas of pin and thrum flowers of Menyanthes trifoliata from population BL3 72 12. Expected and observed pollen frequencies on stigmas of pin and thrum flowers of Menyanthes trifoliata from population P C I 73 13. Expected and observed pollen frequencies on stigmas of pin and homostyle flowers of Menyanthes trifoliata from population SL1 74 14. Seed set and expected seed set in pins, thrums and homostyles of Menyanthes trifoliata from six populations 89 15. Seed set, percentage fruit set and percentage seed set in pins, thrums and homostyles of Menyanthes trifoliata from nine populations 99 v LIST OF FIGURES 1. Typical habitat of Menyanthes trifoliata 7 2. Emergence of an inflorescence of Menyanthes trifoliata. 9 3. Distribution map of Menyanthes trifoliata in southwestern British Columbia. .. 12 4. Longitudinal sections of pin and thrum flowers of Menyanthes trifoliata 17 5. Longitudinal sections of pin, thrum and homostyle flowers of Menyanthes trifoliata 18 6. Pin and thrum plants of Menyanthes trifoliata growing in a growth chamber. '. 32 7. The effects of pollination intensity on seed set in pins and thrums of Menyanthes trifoliata 36 8. Populations of Menyanthes trifoliata at Beaver Lake in Stanley Park, Vancouver, B.C 47 9. Populations of Menyanthes trifoliata located in the Pinecrest area, 0.9 km north of Daisy Lake canal on B.C. Hwy 99 48 10. Populations of Menyanthes trifoliata at Stump Lake in Alice Lake Provincial Park 49 11. Populations of Menyanthes trifoliata located in small ponds in the Pinecrest area 50 12. The percentage of compatible pollen deposited on stigmas of pins, thrums and homostyles in six populations of Menyanthes trifoliata plotted against the percentage of compatible pollen produced in each population 60 13. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL2 65 14. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL3 66 15. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL4 67 16. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population P C I 68 17. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population PC2 69 vi 18. The number of pollen grains deposited on stigmas of pin and homostyle flowers of Menyanthes trifoliata in population S L l 70 19. Seedlings of Menyanthes trifoliata germinating in the mud in ponds in the Pinecrest area 77 20. Populations of Menyanthes trifoliata at Strachan Meadow in Cypress Provincial Park in West Vancouver, B.C 86 21. The number of seeds set per capsule in pins, thrums and homostyles of Menyanthes trifoliata plotted against the number of compatible pollen grains deposited on stigmas 88 22. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL2 92 23. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL3 93 24. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL4 94 25. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population P C I 95 26. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population PC2 96 27. The number of seeds set per capsule in pins and homostyles of Menyanthes trifoliata in population S L l 97 vii ACKNOWLEDGEMENT I thank my major professor, Dr. Fred Ganders for inspiration, advice and financial support. Special thanks are due to Ms. Sue Dreier, Dr. Helen Kennedy and Dr. Jack Maze, and to my committee members Dr. Gary Bradfield and Dr. Wilf Schofield. I am also grateful to my husband John and my family for endless patience and love. viii To J.A.C. who rescued me from countless muddy abysses. i x I. INTRODUCTION As reviewed by Ganders (1979), distyly is a relatively rare plant breeding system in which a single locus, diallelic self-incompatibility system is genetically linked with certain floral dimorphisms. This results in populations of plants consisting of two morphologically distinguishable mating groups in which stigmas and anthers are reciprocally positioned in the two floral forms. One type has flowers with long styles and short stamens and is called the "pin" form. The other type has flowers with short styles and long stamens and is called the "thrum" form. Pollen from pin anthers is compatible with only thrum stigmas, while pollen from thrum anthers is compatible with only pin stigmas. Other floral dimorphisms often associated with the style/stamen dimorphism are differences in pollen size and morphology, pollen production and length of stigmatic papillae. Those distylous species in which pin and thrum pollen are visibly distinguishable provide a unique opportunity for the study of pollen flow in self-incompatible plants. Distyly is known to occur in 25 families of flowering plants (Ganders 1979, Neisess 1984), but was first reported in 1583 by a Flemish botanist named Clusius in studies of Primula (vanDijk 1943). Darwin (1862), was the first to establish the relationship between the floral dimorphisms and the incompatibility system of distylous plants. Through a series of experimental pollinations using Primula, he was able to show that fertilization and seed set would result only from pollinations between flowers of the opposite form (legitimate pollinations). Darwin (1877) suggested that the reciprocal placement of 1 INTRODUCTION / 2 stigmas and anthers in the two floral forms increased the amount of compatible pollen deposited on stigmas by pollinators since the portion of the pollinator's body where pollen from one form was deposited would correspond to the position of the stigma of the opposite floral form. This would result in phenotypic disassortative pollination between pins and thrums. Darwin (1877) hypothesized that the adaptive significance of distyly was to reduce pollen wastage leaving metabolites for other purposes. More recent studies of pollen flow in distylous plants have shown that most of the pollen produced in distylous plant populations is wasted (Ornduff 1970a, 1970b, 1971, 1976, 1979, 1980a; Ganders 1974; Weller 1980; Philipp and Schou 1981; Rama Swamy and Bir Bahadur 1984; Nicholls 1985). An average of 1% of the pollen produced in populations was deposited on conspecific stigmas in these studies. Contrary to Darwin's hypothesis, these studies indicate that distyly may not be an effective mechanism for reducing pollen wastage. Examination of the composition of stigmatic pollen loads of distylous plants has shown that most of the pollen that was deposited on stigmas was incompatible (Levin 1968; Ornduff 1975, 1978; Ganders 1979; Oleson 1979; Lewis 1982; Schou 1983), but this is not surprising given the type of incompatibility system present. Most self-incompatible plants have a multi-allelic incompatiblity mechanism, with several mating groups in each population. But with the diallelic self-incompatibility system found in distylous plants, populations have only two mating groups. With complete self-incompatibility and no differences between morphs in vegetative reproduction or survivorship, the equilibrium frequency of pins and thrums is 1:1 and is reached in one generation (Ganders 1979). With INTRODUCTION / 3 a 1:1 ratio, pollen from one-half of the population is incompatible with any given stigma (Ganders 1979). In assessing the adaptive significance of disytly, Ganders (1974) suggested that more emphasis should be placed on the pollen received by the female, rather than the pollen wasted by the male. Since the chances of the stigma of a diallelic self-incompatible plant receiving incompatible pollen are so great, any mechanism that would increase the number of compatible pollen grains received would increase the fecundit3' of the plant. Because of the reciprocal placement of stigmas and anthers in flowers of distylous plants, it is possible that they would receive a greater proportion of compatible pollen than a diallelic self-incompatible plant lacking the floral dimorphism. Ganders (1974) suggested that the adaptive significance of distyty is to increase fecundity by increasing the number of compatible pollinations received. Ganders' hypothesis assumes that seed set in distylous plants is limited by a lack of compatible pollinations. As reviewed by Shore and Barrett (1984), many factors have been shown to influence seed set in self-incompatible plants. These include: the quantity and efficiency of pollinators present; the density, spatial pattern, and number of mating groups present in the population; the environmental conditions during pollination, fertilization and embryo development; resource availability for fruit and seed maturation; and pollen availability. The influence of these factors on seed set in distylous plants is not known. Since stigmas of distylous plants often receive a small fraction of the pollen produced in the population, and because their chances of receiving incompatible pollen are INTRODUCTION / 4 so great, it is plausible that a lack of compatible pollinations may limit their seed set. The purpose of the present study was to determine the effect of variation in style length on pollen flow in the distylous perennial Menyanthes trifoliata L., and to investigate the relationship between the number of compatible pollen grains received and the number of seeds set. Seed set data were collected in conjunction with pollen flow data, in order to determine whether pollen availability was a major factor influencing seed set in Menyanthes. The influence of variation in morph frequency on pollen flow in populations of Menyanthes was also examined. In most populations of Menyanthes, stigmas and anthers are reciprocally positioned in the two floral forms (pins and thrums). The occurrence of an unusual population of Menyanthes consisting of pins and morphological long homostyles (thrums with styles the same length as their own stamens) provided an opportunity to test Darwin's (1877) theory that morphological distyly increases the proportion of compatible pollen received by the stigmas of diallelic self-incompatible plants. By comparing the stigmatic pollen loads of homostyles to those of thrums from other populations, it was possible to test whether the reciprocal position of stigmas and anthers had any effect on the proportion of compatible pollen received. If, as Ganders (1974) assumes, seed set does correspond to .the number of compatible pollen grains received, and if morphological distyly does increase the proportion of compatible pollen received by stigmas of diallelic self-incompatible plants, then this should result in higher seed INTRODUCTION / 5 set in thrums than in self-incompatible homostyles. II. FLORAL MORPHOLOGY A. INTRODUCTION Menyanthes trifoliata is a distylous, aquatic perennial herb, with a circumpolar distribution between 40 °N and the Arctic circle, and north of the Arctic circle into Greenland, Norway, Alaska and Siberia (Hewett 1964). It grows at elevations ranging from sea level to 3200 meters and inhabits bogs, marshes, lakes and rivers (Figure 1) (Hewett 1964). Water levels in its habitat can range from a few centimeters (often drying out by the end of summer) to depths of 2 meters or more. Menyanthes has submerged horizontal stems with adventitious roots arising at the nodes. The leaves are emergent, with sheathing petioles and ternate, with obovate leaflets 4-12 cm long (Hitchcock and Cronquist 1973). The leaves die back each winter and growth resumes each spring from axillary buds. The inflorescence is a terminal, leafless raceme with up to 40 flowers (Hewett 1964). It is formed at the end of each growing season and remains dormant over winter (Hewett 1964). The flowers have a fused calyx with five ovate lobes and a sympetalous, campanulate corolla 15-30 mm in diameter (Fernald 1950). The corolla is white with pinkish tips and covered with white hairs along the inner surface of the limb. There are five epipetalous stamens, alternate with the corolla lobes. The filaments are white and the anthers are purplish brown, sagittate and contain bright orange-yellow pollen. The pistil is composed of two or three carpels and has a unilocular ovary with parietal placentation. 6 FLORAL MORPHOLOGY / 7 FIGURE 1. Typical habitat of Menyanthes trifoliata. A small pond located nearby Stump Lake in Alice Lake Provincial Park, southwestern British Columbia. F L O R A L M O R P H O L O G Y / 8 In populations of Menyanthes trifoliata found in southwestern British Columbia, growth of leaves and the inflorescence begins in February at low elevations, and in June at high elevations. Initially, one or two leaves elongate, become emergent and expand, followed by the elongation and emergence of the inflorescence (Figure 2). Flowering sometimes begins as early as Apri l , but most often occurs during the month of May (July at higher elevations). The fruit is a round capsule approximately 1 cm in diameter, has a persistent style and dehisces along one or two sutures (Hewett 1964). Seeds may be dispersed by capsule dehiscence or may be released as the infructescence withers, falls in the water and rots. In one population (Beaver Lake) used for this study Red-wing Blackbirds were observed pecking at the fruits and eating the seeds within. Menyanthes is a monotypic genus in the Menyanthaceae. Distyly occurs in three other genera in this family: Fauria, Nymphoides, and Villarsia (Ganders 1979). Little work has been done on the distylous members of this family; Ornduff (1966, 1982, 1986) and Barrett (1980) have studied Nymphoides and Villarsia, and Ganders (1979) has presented some data on Fauria crista-galli and Menyanthes trifoliata. Distyly in Menyanthes was first studied by Darwin (1877). The population described by Darwin was one that consisted of an equal number of pins and thrums. Although some variation in floral characters was found, the flowers of Menyanthes exhibited the same dimorphisms found in other distylous species. Pin pistils were 1.5 times longer than thrum pistils, while thrum stamens were 2 times longer than pin stamens. Pin stigmas were generally F L O R A L M O R P H O L O G Y / 9 F IGURE 2. Emergence of an inflorescence of Menyanthes trifoliata. F L O R A L M O R P H O L O G Y / 10 larger than thrum stigmas, and thrum anthers were larger than pin anthers. Although exact sizes for the pollen grains were not given, the diameter of thrum pollen was found to average 1.2 times larger than pin pollen. Distyly in Menyanthes was next discussed by Warming (1886) who reported that both distylous and homostylous forms occurred in Greenland. This report has since been misquoted, first by Knuth in 1899 and then by Baker in 1959; both stated that Menyanthes was distylous throughout the world, except in Greenland where it was exclusive^ homostylous. Knuth (1899) examined some distylous flowers of Menyanthes, and noted that the stigmatic papillae were longer in pins, and that thrum pollen was larger than pin pollen. Pollen grain sizes were given as 26x50 um for pin, and 63x120 Mm for thrum. Since the turn of the century, the onlj' mention of distyly in Menyanthes has been a report that pins of Menyanthes produced 1.34x more pollen per anther than did thrums (Ganders 1979). Other than these few reports concerning some of the floral dimorphisms in Menyanthes trifoliata, there has been no comprehensive study of distyly in this genus. This chapter presents data that quantify differences between pins and thrums in floral characters associated with the distylous breeding system of Menyanthes trifoliata. Lnterpopulational variation in these characters was also examined. B. MATERIALS AND METHODS Populations of Menyanthes trifoliata were studied for floral morphology in Beaver Lake in Stanley Park, Vancouver, British Columbia, in Stump Lake in F L O R A L M O R P H O L O G Y / 11 Alice Lake Provincial Park, 13 km north of Squamish, on British Columbia Highway 99, and in ponds at Pinecrest, 0.9 km north of Daisy Lake Canal, on British Columbia Highway 99 (Figure 3). In each population 25-30 flowers of each form were collected for morphological measurements. A total of 11 different floral characters were measured (Table 1). Petal length, stamen length and style length were measured from the base of the ovary to the tip of the structure. Pollen production, pollen removal and pollen diameter were estimated from pollen samples that were acetolyzed and then suspended in a 1:1 mixture of lactophenol and glycerin (Erdtman 1960). Collections were made from each population and samples of pin and thrum pollen were treated separately. Pollen production was measured from undehisced anthers collected just before flower buds opened. Approximately 40 flower buds of each form were collected and their anthers pooled. The number of pollen grains present was counted using a haemocytometer, and 20 replicates were counted for each collection (Lloyd 1965). In order to determine how much pollen was removed . from the anthers of Menyanthes flowers by pollinators, shrivelled anthers were collected from approximately 40 flowers of each form just before flowers wilted. The number of pollen grains remaining in the anthers was counted using a haemocytometer (20 replicates) and this number subtracted from corresponding pollen production numbers (Ornduff 1980a, 1980b). Pollen diameter was determined using collections of ten anthers from open flowers of each form. The longest diameter of 300 pollen grains of each form, in each population, was measured using a Vickers M-22 microscope with an image splitting module (Ganders 1976). Ovule number was counted from several hundred ovaries of each F L O R A L MORPHOLOGY / 12 F IGURE 3. Distribution of Menyanthes trifoliata in southwestern British Columbia. Pie diagrams represent estimates of the frequency of floral morphs in each population. Each circle represents a separate population found at each locality. In some cases, populations were located in the same lake (localities: 6, 8, and 9), and in others populations were found in separate lakes or ponds (localities: 2 and 5). The frequency of each floral morph is represented as a portion of the pie diagram. The frequency of pins in each population is represented in white, the frequency of thrums in black, and the frequency of homostyles in stripes. Morph frequencies were estimated from nearest-neighbour counts in some populations (2, 5, 7, 9 and 10), or from counting a representative sample of plants in the population (3, 6 and 8). No estimate of morph frequencies were available for populations 1 or 4. O P i n s 0 thrums ^ homosty les 1 Whistler Swamp 2 P i nec res t PC1 2 3 3 S tan ley Lake 4 Roadside ponds 11.2 km N of B r o h m Lake 5 S t u m p Lake SL1 2 6 Brown ing Lake Q • 7 S t rachan M e a d o w , Cypress Mtn . # SM1 8 H o l l y b u r n Mtn . ' o • o • • Blue Lost L. F i rs t Gentian L. L. 9 B e a v e r Lake 0 9 • BL2 3 4 10 Ladner M a r s h FLORAL MORPHOLOGY / 13 form from each population. The average number of flowers per inflorescence was counted from approximately 200 inflorescences of each form from each population. Data were analyzed using the non-parametric Median test to determine whether significant differences existed between forms in the different characters measured. Tests of significance with ANOVA could not be used because the data did not conform to the underlying assumptions of equality of variance and normality. However, a nested ANOVA was used to partition variation observed in floral characters into three sources: differences between forms, differences between populations and differences between individuals (the residual factor) (Sokal and Rohlf 1981). C. RESULTS Results from measurements of floral characters showed that in two out of three populations sampled (Beaver Lake and Pinecrest) the flowers of Menyanthes trifoliata were distylous (Table 1), but that many of the floral characters displayed a considerable amount of variation at the populational and individual level (Table 2). The third population (Stump Lake), consisted of pins and (morphological) long homostyles. Up to 85% of the variation in style and stamen lengths was accounted for by differences between pins and thrums (Table 2). At Beaver Lake and Pinecrest, anthers and stigmas were reciprocal!}' positioned in the two floral TABLE 1. F lora l characters of p ins , thrums and homostyles of Menyanthes trifoliata from three populat ions: Beaver Lake, Pinecrest and Stump Lake. Values are the mean and standard dev ia t ion ( in parentheses) . Sample s i zes are given Character F1ora1 morph Beaver Lake P i necies t Stump Lake Petal length p i n 16. 1 ( 1.02) 16.2 (1.21) 12.5 (0.92) (mtn) thrum 14.1 (1.13) 15.8 (0.83) 12.8 (0.85)* Stamen length p i n 9.7 (0.56) 9.1 (0.87) 7 . 4 (0.76) (mm) thrum 13.2 (1.12) 13.9 (0.64) 11.7 (0.56)* S t y l e length p i n 15.4 (0.81) 13.2 (1.21) 11.5 ( 1 . 18) (mm) thrum 8.5 ( 1.09) 8.2 (0.53) 9 . 9 (0.64)' D i f f e r e n c e between stamen and s t y l e p i n 5.7 (0.78) 4.1 (0.77) 4.0 (0.72) length (mm) thrum 4.7 (0.81) 5.8 (0.58) 1 . 8 (0.61)• Anther length p i n 2.0 2.5 (0.14) 2 . 1 (0.09) (mm) thrum 2.1 (0.22) 3.0 (0.25) 2.5 (0.11)* P o l l e n production p i n 17,408 (6,309) 15,539 (1,838) 17,783 (6,604) per f1ower thrum 10.617 (3.268) 18.095 (1,753) 9,990 (1.754)* Po11 en d i ameter p i n 4 1.3 (2.90) 4 1.6 (2.80) 4 1.7 (2.30) (pm) thrum 47.8 (3.17) 50. 1 (3. 10) 47 . 4 (2.90)* Ovary length p i n 3.4 (0.44') 2.5 (0.41) 2 . 4 (0.36) (mm) thrum 2.9 (0.39) 3.0 (0.21) 2.9 (0.22)* Ovu1e number p i n 37.4 (13.38) 30. 5 (6.81) 17.2 (5.93) per f1ower thrum 30.2 ( 12.88 ) 29.3 (8.52 ) 20. 1 (5.43)* Number of p i n 2.3 (0.47) 2.6 (0.49) 2 . 7 (0.48) s t i gma 1obes thrum 2.0 (0.19) 2.3 (0.45) 2.0* Number of flowers p i n 20.6 (3.62) 15.1 < 2.87) 28. 5 (3.46) per i n f l o r e s c e n c e thrum 21.3 (4.56) 15.2 (3.09) 27 . 2 (3.94)* * homostyle TABLE 2. V a r i a n c e components of v a r i a t i o n which i s a t t r i b u t a b l e f l o r a l c h a r a c t e r s i n Menyanthes to i n t e r p o p u l a t i o n a l v a r i a t i o n . trifoliata. V a l u e s a r e the p e r c e n t a g e of o b s e r v e d Intermorph v a r i a t i o n and I n d i v i d u a l v a r i a t i o n . C h a r a c t e r P o p u l a t i o n Morph I n d i v i d u a l P e t a l l e n g t h Stamen l e n g t h S t y l e l e n g t h Anther l e n g t h P o l l e n p r o d u c t i o n P o l l e n diameter Ovary l e n g t h Ovule number Number of Stigma lobes Number of f l o w e r s per i n f l o r e s c e n c e 51 14 60 36 19 38 10 76 85 22 53 58 27 4 25 62 39 10 12 18 1 1 40 54 58 58 36 r o 3 o to X o tr1 O O F L O R A L M O R P H O L O G Y / 16 forms (Figure 4). The difference in length between stamens and styles averaged 4.9 mm (s.d. = 1.13) in pin flowers, and 5.2 mm (s.d. = 0.88) in thrum flowers. On average, the stigmas of one floral form were at the same level as the middle of the anther sacs of the opposite floral form (Table 1). At Stump Lake the difference in length between stamens and styles in pin flowers averaged 4.0 mm (s.d. = 0.72), similar to that found in pin flowers from the other populations. Morphologically, the rest of the flowers from Stump Lake appeared to be long homostyles. The stamens of the homostyle flowers were the same length as the pin styles, but only an average of 1.8 mm (s.d. = 0.61) longer than their own styles (Figure 5). This placed homostyle stigmas at the same level as the middle of their own anther sacs, and an average of 2.5 mm higher than pin stamens. Because homostyle stamens were the same length as pin styles, and homostyle styles were considerably longer than pin stamens, style length, not stamen length, was thought to be the variable factor. The homostyles had longer styles than normal thrums. A t Beaver Lake and Pinecrest, pin pollen was significantly (p< 0.001) smaller in diameter than thrum pollen, while at Stump Lake, pin pollen was significantly (p< 0.001) smaller in diameter than homostyle pollen. The longest diameter of pin pollen grains averaged 41.5 um (s.d = 2.69), while the longest diameter of thrum pollen averaged 48.9 um (s.d. = 3.27). In each population the size ranges for the two pollen types overlapped slightly (Table 1). Between populations, average pollen diameters were similar except for thrum pollen from Pinecrest, which was on average 2.3 um larger in diameter than thrum pollen F L O R A L MORPHOLOGY / 17 F IGURE 4 . Longitudinal sections of pin and thrum flowers of Menyanthes trifoliata from Pinecrest in southwestern British Columbia showing reciprocal placement of stigmas and anthers in the two floral forms. FIGURE 5. L o n g i t u d i n a l s e c t i o n s of p i n , homostyle and thrum f l o w e r s of Menyanthes trifoliata. F L O R A L M O R P H O L O G Y / 19 from Beaver Lake (Table 1). Pollen from homostyles at Stump Lake was similar in size (47.4 um, s.d. = 2.90) to thrum pollen from Beaver Lake (47.8 um, s.d. = 3.17). A total of 53% of the variation in pollen production was accounted for by differences between morphs, but there was also a considerable amount of variation i n pollen production between populations (36%) (Table 2). Pin flowers produced more pollen than thrum flowers at Beaver Lake and homostyle flowers at Stump Lake, but at Pinecrest, thrum flowers produced more pollen than pin flowers (Table 1). At Pinecrest, thrum flowers had 1.65 times more pollen grains removed by pollinators than did pin flowers. At Beaver Lake, pin flowers had 1.67 times more of their pollen removed than did thrum flowers (Table 3). Within each population, thrum anthers (and homostyle anthers) were significantly longer than pin anthers (Stump Lake and Pinecrest p<0.001, Beaver Lake p<0.05), but the variation among populations (60%) was greater than the variation between morphs (22%) (Table 2). Pin and thrum anthers from Pinecrest were the largest of the three populations. Anther length averaged 2.5 mm (s.d. = 0.14) in pins, and 3.0 mm (s.d. = 0.25) i n thrums. Anthers from Beaver Lake were the smallest where anther length in pins was 2.0 mm, and averaged 2.1 mm (s.d. = 0.22) in thrums. Petal length also varied more among populations than among individuals or between morphs (Table 2). Petals from Pinecrest were the longest, while those from Stump Lake were the shortest. Beaver Lake was the only population where TABLE 3. Number and percentage of pollen grains removed from pin and thrum flowers of Menyanthes trifoliata by p o l l i n a t o r s . Values are the mean and standard deviation ( i n parentheses), and represent the amount of pollen removed from one flower ( f i v e anthers per flower). Population Floral Morph Average number Average percentage of pollen grains of pollen grains removed from flower removed from flower Beaver Lake pin thrum 41 GO 13,396 (9,257) 8,037 (3,022) 64.7 ( 18.47) 70.8 (8.15) Pinecrest pin thrum 23 18 10.14 1 (3,300) 16,732 (1,091) 62.8 (14.00) 90.4 (3.88) o so > O so X O o o t o o F L O R A L M O R P H O L O G Y / 21 there was a significant difference (p< 0.001) in petal size between pins and thrums; pin petal length averaged 16.1 mm (s.d. = 1.02), whereas thrum petal length was only 14.1 mm (s.d. = 1.13). A t Pinecrest, petal length in pins averaged 16.2 mm (s.d. = 1.21), and in thrums 15.8 mm long (s.d. = 0.83). Pin petals from Stump Lake averaged 12.5 mm in length (s.d. = 0.92), while homostyle petal length averaged 12.8 mm (s.d = 0.85). The remaining three floral characters, ovary length, ovule number and the number of stigma lobes were most variable at the individual level (Table 2). Ovary length varied significantly between pins and thrums and pins and homostyles within populations, but this variation was not consistent among populations (Table 1). Homostyle ovaries were larger than pin ovaries in flowers from Stump Lake (pin: 2.4 mm, s.d. = 0.36; homostyle: 2.9 mm, s.d. = 0.22; p<0.001). At Pinecrest, thrum ovaries were significantly larger than pin ovaries (pin: 2.5 mm, s.d. = 0.41; thrum: 3.0 mm, s.d. = 0.21; p<0.001), but the reverse was true at Beaver Lake (pin: 3.4 mm, s.d. = 0.44; thrum: 2.9 mm, s.d. = 0.39; p<0.001). Ovule number was significantly different between pins and homostyles at Stump Lake (p<0.001), and pins and thrums at Beaver Lake (p<0.001), but there was no significant difference in ovule number between forms at Pinecrest. In those populations where there was a significant difference between forms in ovule number, this difference corresponded to differences in ovary length. At Stump Lake, ovary length was greater for homostyles and ovule number averaged 17.2 (s.d. = 5.93) for pins, and 20.1 (s.d. = 5.43) for homostyles. At Beaver Lake, ovary length was greater for pins, and ovule number averaged 37.4 (s.d. = 3 3.38) for pins, and 30.2 (s.d. = 12.88) for thrums. There was no F L O R A L M O R P H O L O G Y / 22 significant difference between pins and thrums or between populations in stigma lobe number. While the number of flowers per inflorescence did not vary significantly between morphs, a total of 62% of the variation in the number of flowers per inflorescence was accounted for by differences between populations (Table 2). Plants at Stump Lake had the most flowers per inflorescence, for pins this number averaged 28.5 (s.d. = 3.46), and for homostyles 27.2 (s.d. = 3.94). Plants from Pinecrest had the fewest, for pins this number averaged 15.1 (s.d. = 2.87), and for thrums 15.2 (s.d. = 3.09). In general, it appeared that inflorescences from Pinecrest had large flowers, few in number, while those from Stump Lake had smaller but more numerous flowers. D. DISCUSSION Distyly is known to occur in 24 families of flowering plants (Ganders 1979). Besides the reciprocal placement of anthers and stigmas, flowers of distylous plants have several other morphological characters in common. Flowers of Menyanthes trifoliata share some, but not all of these characteristics. As is commonly found in distylous plants, corollas of Menyanthes are small, regular, salverform and sympetalous, with a short floral tube (Ganders 1979). In some distylous species, thrums have a larger corolla than pins (Ganders 1979), but in Menyanthes, petal length varied more between populations than between forms. In the one population (Beaver Lake) where there was a significant difference between the petal length of pins and thrums, the pins had longer petals. The F L O R A L M O R P H O L O G Y / 23 present study confirms Darwin's (1877) report that in Menyanthes thrum anthers are larger than pin anthers. Larger thrum anther size has been reported in several other distylous species including other members of the Menyanthaceae (Ornduff 1966, 1982; Ganders 1979). However, in Menyanthes interpopulational variation in anther size is greater than intermorph variation. The pollen size dimorphism in Menyanthes was consistent with that found in other distylous species; pin pollen was smaller in diameter than thrum pollen. The pollen diameters for pins and thrums of Menyanthes determined in the present study (pin = 41.5 urn, s.d. = 2.69; thrum = 48.9 um, s.d. = 3.27) were similar to those given by Nilsson (1973) (pin = 33.5 jum, s.d. = 0.54; thrum = 40.4 Mm, s.d. = 0.57). In agreement with Darwin's (1877) findings, results of the present study showed that thrum pollen was an average of 1.2 times larger in diameter than pin pollen. This is in contrast to the much larger sizes reported by Knuth (1899) and later quoted by Hewett (1964) (pin = 26x50 um, thrum = 63xl20 Mm). According to Knuth's measurements, thrum pollen was 2.4 times larger than pin pollen. Reasons for the differences between Knuth's (1899) measurements of pollen sizes and those found in the present study are not known. In Menyanthes, the number of pollen grains produced by pins and thrums varied considerably among populations. The ratio of pin pollen production to thrum pollen production at Beaver Lake was 1.64 while at Pinecrest the ratio was 0.86. At Stump Lake, the ratio of pin pollen production to homostyle pollen production was 1.78. In most distylous species studied, pins produced more pollen than thrums, and this difference is correlated with a difference in pollen size F L O R A L M O R P H O L O G Y / 24 (Ganders 1979). Linum perenne (Nicholls 1986) and two species of Amsinckia (Ganders 1979) are the only known distylous taxa besides Menyanthes in which higher thrum pollen production has been reported. Because of the interpopulational variation in pollen production in Menyanthes it is difficult to correlate differences in pollen production between pins and thrums with any developmental features. In Menyanthes, variation in pollen production does not correspond well to variation in anther size or pollen size. In those distylous taxa in which pollen removal by pollinators has been measured, a greater number of pollen grains were removed from thrum anthers than pins anthers (Ornduff 1980a, 1980b). Ornduff suggested that this was because thrum anthers were more accessible to pollinators than pin anthers. In Menyanthes, the number of pollen grains removed from pin and thrum anthers varied considerably between populations. A t Pinecrest, thrum flowers had 1.65 times more pollen grains removed than did pin flowers, but at Beaver Lake, pin flowers had 1.67 times more of their pollen removed than did thrum flowers. So that in Menyanthes, anther accessibility may not have been a factor influencing the number of pollen grains removed by pollinators. Intermorph variation in pollen removal did correspond to intermorph variation in pollen production. At Beaver Lake and Pinecrest, the morph which produced more pollen also had more pollen removed by pollinators. While flowers of Menyanthes trifoliata exhibited many of the floral dimorphisms characteristic of the distylous breeding system, in ' some of these characters variation between populations was greater than variation between FLORAL MORPHOLOGY / 25 morphs. By far the most significant example of interpopulational variation in floral morphology in Menyanthes is the variation in style length found in flowers from Stump Lake. At Beaver Lake and Pinecrest, there is a strong reciprocal relationship between style length and stamen length in pins and thrums. In these two populations, stigmas of one morph are at the same level as the middle of the anther sacs of the opposite morph. While pins from Stump Lake exhibit a similar degree of separation of their stigmas and anthers as is found in flowers from other populations, the rest of the flowers from Stump Lake appear to be long homostyles. Their stamens are the same length as the pin styles, but the homostyle styles are longer than those expected in the thrum form. Homostyle stigmas are at the same level as their own anther sacs. Interpopulational variation in style length (and stamen length) has also been found in thrums from three species of Villarsia, another distylous genus in the Menyanthaceae (Ornduff 1986). Three populations of V. parnassiifolia, two of V. capitata and two of V. lasiosperma were studied, and in each of these populations pins appeared normal. In six of these populations, stamens of the other floral form (referred to as "thrums" by Ornduff) were shorter than the pin styles and at the same level as or longer than pin stamens. Style length of these thrums varied amoung populations from above the distal tip of their own stamens, to below the proximal tip. In four populations (three of V. parnassiifolia and one of V. capitata), the thrums appeared to be short homostyles. Only in one population (V. lasiosperma), was there a reciprocal relationship between style and stamen length in the two floral forms. Ornduff (1986) concluded that in the populations of Villarsia that he studied, thrum flowers were highly variable in FLORAL MORPHOLOGY / 26 both style and stamen length. He suggested that those flowers that appeared homostylous were merely a result of morphological variation, but offered no evidence to support this. As reviewed by Ganders (1979), most homostyles in distylous species are a result of a crossover in the supergene controlling the compatibility reaction and the stamen-style length polymorphism. Such a crossover results in flowers with the pistil of one form and the stamens of the opposite form, so that true genetic homostyles are self-compatible. Self-incompatible homostyles are very rare, but two examples have been reported. In both cases, the homostylous appearance has been attributed to morphological variation, either elongated pin stamens {Primula farinosa) or elongated thrum styles (Mitchella repens). In order to determine whether a homostyle is a result of genetic recombination or morphological variation, one must self-pollinate the homostyle to test for self-compatibility. Unfortunately this has not been done for Menyanthes homostyles from Stump Lake. However as in Villarsia, flowers of Menyanthes seem to have a great capacity for morphological variation. In subsequent chapters, evidence is presented to support the view that the homostyles at Stump Lake are not true genetic homostyles, but are a result of morphological variation, that is, they are thrums with elongated styles. The discovery of a homostylous form of Menyanthes at Stump Lake is the first report of homostyly in this species since Warming's (1886) report of homostyles in Greenland. Whether the homostyles found by Warming were true FLORAL MORPHOLOGY / 27 genetic homost^ des, or simply a result of morphological variation is not known. However the morphological variation in style length in flowers from Stump Lake is a significant discovery, because it provides an opportunity to test Darwin's (1877) theory that morphological distyly increases the proportion of compatible pollen received by a diallelic self-incompatible plant. If one assumes that the only difference between the Stump Lake homostyles and thrums from other populations is the length of their styles, any difference in the proportion of compatible pollen received by thrums and homostyles must be a result of the variation in their style length. III. EXPERIMENTAL POLLINATIONS A. INTRODUCTION This chapter describes the results from two types of experimental pollination programs using Menyanthes trifoliata plants grown in a growth chamber. The first was to determine whether a self-incompatibility system is associated with morphological distyly in Menyanthes trifoliata. The second was to quantify the relationship between the number of compatible, pollen grains deposited on a stigma (pollination intensity) and the number of seeds set. Distylous plants typically have a relatively inefficient self-incompatibility mechanism controlled by two alleles at one locus, which is genetically linked with the loci controlling the floral pofymorphisms (Ganders 1979). The strength of the incompatibilty reaction varies in different distylous species, and even some self-compatible taxa are known: some species of Oxalis and Hedyotis, Melochia pyramidata; and all species of Amsinckia. More recently, self-compatibility has been reported in distylous Salvia brandegei (Neisess 1984) and Cryptantha flava (Casper 1985). Data on self-incompatibility systems are available for only two of the four distylous genera in the Menyanthaceae: Nymphoides (Ornduff 1966; Barrett 1980) and Villarsia (Ornduff 1982). Controlled crossing programs have shown that Nymphoides indica (Ornduff 1966; Barrett 1980), Nymphoides humboldtiana (Ornduff 1966) and Villarsia capitata (Ornduff 1982) have the same diallelic 28 E X P E R I M E N T A L POLL INATIONS / 29 self-incompatibility system found in other distylous plants. In Nymphoides peltata the self-incompatibility system was weak, and illegitimate crosses resulted in substantial seed set (Ornduff 1966). To date, no one has tested whether the morphological characteristics of distyly in Menyanthes are associated with a self-incompatibility system. The only evidence of self-incompatibility in Menyanthes comes from Darwin's (1877) report of a population of Menyanthes which consisted entirely of thrums. In this population, the capsules expanded, but they did not contain any viable seed. Darwin (1877) suggested that this was an indication that Menyanthes was self-incompatible. In the present study a experimental pollination program was designed to test for the presence of a self-incompatibility system in Menyanthes trifoliata. In order to further our understanding of breeding systems in plants, another aspect of the pollination/fertilization process warrants investigation. The relationship between pollination intensity (the number of compatible pollen grains deposited on a stigma) and seed set is an aspect of the reproductive biology of plants that has been virtually ignored. Only recently have attempts been made to quantify this relationship (Akamine and Girolami 1959; Silander and Primack 1978; Bertin 1978; Snow 1982; McDade 1983; Schemske and Fenster 1983; Shore and Barrett 1984; Guth and Weller 1986). These studies indicate that there may not always be a 1:1 relationship between the number of compatible pollen grains received and the number of seeds set, and this relationship may vary among species. For the best results, the relationship between pollination intensity and seed E X P E R I M E N T A L POLL INATIONS / 30 set should be determined under field conditions. Ideally, seed set should be measured in the same flowers in which the number of compatible pollen grains received has also been determined. McDade (1983) used a hand lens to count the number of pollen grains deposited on stigmas of day old flowers of the non-heterostylous species Tricanthera gigantea then later measured fruit and seed set, but only pollen from stigmas with fewer than eight pollen grains could be reliably counted. This method is not very practical in distylous plants. In order to distinguish between the compatible and incompatible pollen grains deposited on a stigma, the stigma must be removed and its pollen counted under a compound microscope. But once the stigma and its pollen have been removed, so are the chances of effecting fertilization and seed set. An alternative method is to hand pollinate flowers with a known number of compatible pollen grains and then determine seed set. This method was used by Snow (1982) on field grown Passiflora uitifolia, but she was unable to determine the exact number of pollen grains applied to stigmas, so results were only approximate. This method is performed more easily under the controlled conditions of a pollinator free growth chamber or greenhouse, and this was the procedure followed by Shore and Barrett (1984), Guth and Weller (1986) and in the present study. Results from these experimental pollinations provide an estimate of how many compatible pollen grains must be deposited on a stigma in order to fertilize each ovule in Menyanthes trifoliata. These results serve as a basis for an investigation into the relationship between the number of compatible pollen grains received and the number of seeds set in natural populations, of Menyanthes . (Chapter 5). E X P E R I M E N T A L POLL INATIONS / 31 B. MATERIALS AND METHODS Pin and thrum plants of Menyanthes trifoliata transplanted from Beaver Lake, Lost Lake and Pinecrest (Figure 3), were potted and placed in trays full of water in pollinator-free growth chambers (Figure 6). Four different types of experimental pollinations were performed. A total of 276 flowers were pollinated; 118 flowers were selfed with pollen from their own anthers, while 71 flowers were pollinated with pollen from other flowers on the same inflorescence (geitonogamous pollination). Fifty-four intra-morph cross pollinations (between different plants of the same form) and 33 inter-morph cross pollinations (between flowers of different forms) were made. Nine pin plants and four thrum plants were recipients of all four crossing regimes. Pollen was transferred by removing, with jeweller's forceps, freshly dehisced anthers from the pollen donor and rubbing the anther across the recipient stigma until all visible traces of the pollen had been transferred. Fruits were allowed to mature, but harvested before dehiscence in order to count how many seeds had been set. Seventy-eight experimental pollinations were made in which the exact number of compatible pollen grains was determined before transferring them to the recipient stigma. First, a random numbers table was used to select some number of compatible pollen grains between 1 and 99. Then, a dehisced anther from ' a donor plant was removed with jeweller's forceps and brushed against a glass slide to deposit its pollen on the slide. Under a dissecting microscope, a EXPERIMENTAL POLLINATIONS / 32 FIGURE 6. Pin and thrum plants of Menyanthes trifoliata growing in a growth chamber. EXPERIMENTAL POLLINATIONS / 33 needle was used to count out the required number of pollen grains and then to transfer them to the stigma of the recipient plant (Shore and Barrett 1984). Fruits were allowed to mature and seed set was determined. A total of 29 pin flowers (on 12 separate plants), and 35 thrum flowers (on 11 separate plants) were used as pollen recipients. C. RESULTS A considerable amount of variation was noted in the fertility of individual plants of Menyanthes used for both of the experimental pollination programs. Some plants failed to set seed regardless of the amount of compatible pollen applied to their stigmas, and these plants were excluded from the results. Reasons for this variation in fertility are not known. The controlled crossing program provided the first experimental evidence that Menyanthes trifoliata possesses the same type of self-incompatibility system as other distylous plants. While self, geitonogamous and intra-morph crosses resulted in less than 2% seed set and averaged 0.75% (s.d. = 0.78), crosses between pins and thrums alwa3rs resulted in substantial seed production averaging 43.7% (s.d. = 12.9) (Table 4). Observations of seed set in natural rnonomorphic populations of Menyanthes support these results. Two populations which consisted only of thrums located at Strachan Meadow and at Pinecrest, and one composed entirely of morphological homostyles located at Stump Lake (Figure 3), all had seed set lower than 4% TABLE 4. Seed set fo l lowing s e l f , ge1tonogamous, intramorph and Intermorph p o l l i n a t i o n of pins and thrums of Menyanthes trifoliata. Plants were grown from cutt ings c o l l e c t e d from Beaver Lake, Pinecrest and Lost Lake (Figure 3 shows l o c a t i o n s ) . A l l p o l l i n a t i o n s were performed in a growth chamber. Values are the mean and standard dev ia t ion (i n parentheses ) . Type of pol1inat ion Number of flowers pol11nated Average number of seeds set per pol1inat Ion Average percentage seed set per pol11nat1on Intrap 1 ant p in se 1 f 83 0.30 (0.94) 1.30 (4.36) thrum se l f 35 0.03 (0.16) 0.09 (0.53) pin geitonogamous 40 0.47 (1.50) 2.00 (6.16) thrum geitonogamous 4 1 0.09 (0.53) 0.36 (1.99) Interplant Intramorph pin x p in thrum x thrum Intermorph pin x thrum thrum x p in 45 9 23 10 0. 16 (0.56) 0 13.80 (8.95) 8.00 (6.21) 0.73 (2.60) 0 52.80 (29.47) 34.50 (28.50) X W S> H H > *d o r > O w CO E X P E R I M E N T A L POLL INATIONS / 35 (Chapter 5, Table 15). At Strachan Meadows in 1983, the capsules expanded to at least one half the size of mature fruits, but contained only unfertilized ovules. In the same population in 1985, the capsules did not expand to such an extent and were heavily infested with aphids. Average .seed set for 1985 was 0.01 seeds/capsule (s.d. = 0.20). Results from pollination intensity experiments show that in Menyanthes, an increase in the number of compatible pollen grains received by a stigma generally resulted in an increase in the number of seeds set, but there was a considerable amount of variation in seed set at each level of pollination intensity (Figure 7). The number of seeds set was always less than the number of compatible pollen grains received. In nine flowers (on separate plants) very high levels of compatible pollen resulted in very low seed set. For example, in one flower only one ovule was fertilized after 98 compatible pollen grains were deposited on its stigma. These flowers were not excluded from the dataset because other flowers on the same plants had much higher seed fertility, indicating that the plants were not infertile, but were capable of seed production. Inclusion of these flowers in the data set doubled the calculated average number of pollen grains required to set each seed. The average number of pollen grains required to set each seed (including the nine flowers: n = 78) was 13.5 (s.d. = 19.82), compared to an average of 7.7 pollen grains per seed set (s.d = 7.45) when the results from those nine flowers were excluded (n = 69). The median number of pollen grains per seed set gives a more accurate indication of the relationship between pollination FIGURE 7. The e f f e c t s of p o l l i n a t i o n Intensity on seed set in pins and thrums of Menyanthes trifoliata. Individual pin flowers are represented as closed c i r c l e s , thrum flowers as open c i r c l e s . Average number of ovules per ovary is 25.3 (s.d.=7.55). Broken l i n e represents the least squares regression of seed set on p o l l i n a t i o n intensity for a l l datapoints. Correlation: r=0.37, p<0.01. Solid line represents the least squares regression excluding nine datapoints represent 11ng flowers with extremely low seed f e r t i l i t y (see text for explanation). Correlation: r=0.57. p<0.01. 50 60 70 80 Number of Compatible Pollen Gra ins Deposited on St igmas 90 100 > o z CO E X P E R I M E N T A L POLLINATIONS / 37 intensity and seed set in Menyanthes than the mean number. The median number of compatible pollen grains required to fertilize each ovule was 6.0 with the nine low seed set flowers included in the dataset, and 5.0 when they were excluded. The small difference between the two medians indicates that the results from the nine flowers with very low seed set were not characteristic of the whole data set. Results from pins and thrums were similar; for pins 6.0 was the median number (n = 37) of compatible pollen grains required to set each seed, while for thrums the median was 5.3 (n = 41). The number of compatible pollen grains applied to stigmas ranged from 1-99, but the lowest number of compatible pollen grains that resulted in seed set was three. Two linear regressions were performed in order to evaluate the relationship between the number of compatible pollen grains received and the resultant seed set. Regression lines from the two analyses are shown in Figure 7: the broken line represents the regression of seed set on pollination intensity for all data points (n = 73, r = 0.37, p<0.01), while the solid line represents the regression of all data points except the nine flowers that had very low seed set (n = 69, r = 0.57, p<0.01). Results from these analyses show that there is a significant correlation between pollination intensity and seed set. The correlation is not especially high, which underscores the great variation in the results. D. DISCUSSION As in other distylous members of the Menyanthaceae, morphological distyly is associated with a self-incompatibility system in Menyanthes trifoliata. In E X P E R I M E N T A L POLL INATIONS / 38 Menyanthes, seed set levels for self, intra-morph and inter-morph crosses were similar to those found for Nymphoides humboldtiana (Ornduff 1966), Nymphiodes indica (Barrett 1980) and Villarsia capitata (Ornduff 1982). In Menyanthes, crosses using own-form pollen resulted in less than 2% seed set in both pins and thrums. In inter-morph pollinations pins tended to set more seed (13.8 seeds/capsule, s.d. = 8.95) than thrums (8.0 seeds/capsule, s.d. = 6.21). In most distylous species studied, seed set was similar for pins and thrums following legitimate hand-pollinations (Barrett 1980). In a few species, Primula spp. (Dowrick 1956), Eichhornia crassipes (Barrett 1977), Turnera ulmifolia (Barrett 1978), and Nymphoides indica (Barrett 1980), seed set following hand-pollination was lower in pins. Dowrick (1956) attributed lower pin seed set in Primula obconica to the smaller area of conducting tissue available for pollen tube growth. In contrast to this, reports of lower seed set in thrums following hand-pollinations in Nymphoides indica (Reddy and Bahadur 1976), Villarsia capitata (Ornduff 1982) and in Menyanthes trifoliata in the present study indicate that the area of stylar conducting tissue is probably not a major factor influencing seed set in these species. In Menyanthes both pin and thrum seed set following inter-morph pollination was less than the potential maximum. Out of 33 inter-morph pollinations, only eight resulted in greater than 80% seed set. This may be a result of resource availability, if the plants grown in growth chambers were not as healthy as those in natural populations more ovules may have aborted. Variation among individuals in seed and/or pollen fertility may also be a E X P E R I M E N T A L POLL INATIONS / 39 contributing factor. Variation in fertility was found in hand-pollinated Turnera ulmifolia (Shore and Barrett 1984). As in Menyanthes, some plants failed to set seed regardless of how much compatible pollen was applied to their stigmas. Barrett (1980) was able to demonstrate significant difference among individuals of Nymphoides indica in average seed fertility, but no differences in pollen fertility were apparent. More data are necessary before a similar analysis of variation in seed and pollen fertility can be performed in Menyanthes. The observation of expanded capsules in the monomorphic thrum population at Strachan Meadow is very puzzling. Darwin (1877) made similar observations in a monomorphic thrum population located at Kew Gardens. Darwin reported that fruit set (i.e., expanded capsules) occurred without the production of viable seed. In most natural populations of Menyanthes that I observed, fruit set does not occur without seed set, however parthenocarpy does occur in another distylous member of the Menyanthaceae, Fauria crista-galli (Griffiths and Ganders 1983). Why it would occur only in some populations of Menyanthes and not in others, and vary among years is not understood. In Menyanthes, caution should be exercised when fruit set is used as an indication of successful fertilization; fruit set should be measured in conjunction with seed set. Results from the pollination intensity experiments in Menyanthes trifoliata show that there was not a 1:1 relationship between the number of compatible pollinations received and seed set. The number of seeds set was always lower than the number of compatible pollen grains received, and the median number of compatible pollen grains required to fertilize each ovule was 6.0. Similar results E X P E R I M E N T A L POLL INATIONS / 40 were found for distylous Turnera ulmifolia (Shore and Barrett 1984). In T. ulmifolia, there was a considerable amount of variation in the number of seeds set at each level of pollination intensity, but in general, seed set increased as the number of compatible pollinations increased (Shore and Barrett 1984). Seed set was always below the potential maximum, and it took an average of 2-7 compatible pollen grains to fertilize each ovule (Shore and Barrett 1984). Similarily, in Tricanthera gigantea, at least eight pollen grains were necessary for fruit and seed set (McDade 1983). Other studies have shown that for several species {Passiflora edulis (Akamine and Girolami 1959), Oenothera fruticosa (Silander and Primack 1978), Passiflora uitifolia (Snow 1982), and Oxalis magnifica (Guth and Weller 1986)}, seed set increased as the number of compatible pollinations increased, but the efficiency of pollen grains effecting fertilization decreased at higher levels of pollination intensity. In Oenothera fruiticosa there was close to a 1:1 relationship of compatible pollen grains to seed set when the number of pollen grains applied to stigmas was below one third the number required for maximum seed set (Silander and Primack 1978). A similar relationship was found in Oxalis magnifica when pollination levels were below 35 pollen grains per stigma (Guth and Weller 1986). A t higher levels of pollination intensity, more than one pollen grain was required to set each seed (Guth and Weller 1986). Akamine and Girolami (1959) reported a similar decrease in pollen efficiency for Passiflora edulis; at low pollination intensities (189 pollen grains/stigma) an average of 2.2 pollen grains were required for each fertilized ovule, but at higher intensities (1776 pollen grains/stigma) an average of 7.3 pollen grains were required. With E X P E R I M E N T A L POLL INATIONS / 41 an average of 350 ovules per flower, maximum seed set was never achieved even at the highest pollination intensity levels (Akamine and Girolami 1959). In Passiflora vitifolia an average of 1.6 pollen grains were required to set each seed when the number of compatible pollen grains per stigma was less than 100, but more pollen was required at higher pollination intensity levels (Snow 1982). Because so few studies of pollination intensity have been done, it is difficult of make any generalizations concerning the relationship between the number of compatible pollen grains a stigma receives and the number of seed set. Results from the few published studies and from Menyanthes show that the number of compatible pollen grains required for each fertilized ovule varies among species and is frequently not a 1:1 ratio. In some species the relationship between compatible pollen grains and seed set is not linear, the efficiencj' of pollen grains effecting fertilization decreases at higher levels of pollination intensity. This is not particularily surprising, because competition among pollen grains might be expected at higher pollination intensities. More work needs to be done to provide a better understanding of the relationship between pollination intensity and seed set. IV. POLLEN FLOW A. INTRODUCTION In 1877 Darwin hypothesized that distyly should increase the amount of compatible pollen deposited on stigmas because of the reciprocal placement of anthers and stigmas in the two floral forms. He suggested that the portion of the pollinator's body where pollen from one floral morph was deposited would come in contact with the stigma of the opposite floral morph. This would ensure that pins were pollinated with thrum pollen and that thrums were pollinated with pin pollen (disassortative pollination). Few studies of pollen flow in distylous species have been able to support this hypothesis. Studies have shown that stigmas of distylous plants have high levels of incompatible pollen deposited on them, and that pollen flow between pins and thrums is asymmetric (Ganders 1979). In general, it has been found that pin stigmas receive more pollen (pin and thrum pollen combined) than thrum stigmas (Ganders 1979; Ornduff 1980a, 1980b; Weller 1980; Lewis 1982; Casper 1983; Schou 1983; Rama Swamy and Bir Bahadur 1984), that pin stigmas receive more incompatible pollen than compatible pollen (Ganders 1979; Oleson 1979; Ornduff 1980a, 1980b; Weller 1980; Lewis 1982; Schou 1983; Nicholls 1985), and that thrum stigmas receive more compatible than incompatible pollen (Ganders 1979; Ornduff 1980a, 1980b; Lewis 1982; Schou 1983). However, examination of total stigmatic pollen loads is not the most 42 P O L L E N F L O W / 43 effective method of testing whether flowers are disassortatively pollinated (Ganders 1974). This is because the intraflower selfing component of stigmatic pollen loads can mask any disassortative pollen flow by creating an excess of own-form pollen on the stigma. Distyly does not affect the amount of intraflower selfing a flower undergoes, but through disassortative pollination should reduce the amount of own-form pollen derived from other flowers, and increase the amount of opposite-form pollen received. With 100% efficiency and intraflower selfing excluded, all the pollen on stigmas should be of the opposite form. With random pollination (expected if distyly has no effect on pollen flow), the composition of stigmatic pollen loads would be the same as the frequency of pollen produced in the population. The intraflower component of a stigmatic pollen load can be eliminated through emasculation. By comparing observed pollen loads from stigmas of emasculated flowers with expected pollen loads based on random pollination, it is possible to determine whether flowers are being disassortatively pollinated. Ganders (1974) is one of the few researchers who used this method to demonstrate significant levels of disassortative pollination in a distylous plant {Jepsonia heterandra). While Ganders' (1974) study documented the occurrence of disassortative pollination in a distylous species, it did not test whether the disassortative pollination was a result of the stamen/style length dimorphism. However, Ganders (1979) was able to demonstrate that in a population of Lithospermum californicum small differences in the degree of separation between the stigmas and anthers of pin flowers influenced the proportion of compatible pollen they received. Pin flowers with a larger spatial separation between stigmas and anthers received a P O L L E N F L O W / 44 higher percentage of compatible pollen than the other pin flowers. Ideally, the relationship between morphological distyly and disassortative pollination could be tested by comparing the composition of stigmatic pollen loads of distylous flowers, with the composition of pollen loads from otherwise identical diallelic self-incompatible flowers lacking the stamen/style dimorphism. If, as Darwin hypothesized (1877), morphological distyly does increase the amount of compatible pollen received by a diallelic self-incompatible plant, then stigmas of distylous. flowers would be expected to receive a greater proportion of compatible pollen than the stigmas of the homomorphic flowers. The r- occurrence of an unusual population of Menyanthes trifoliata consisting of pins and (morphological) homostyles provided an opportunity to test Darwin's hypothesis. Except for the length of their styles, these homostyles are identical morphologically to thrums from other populations of Menyanthes. By comparing pollen flow in the pin/homostyle population to pollen flow in a similarily structured pin/thrum population, it was possible to investigate the effect of morphological distyly on the composition of stigmatic pollen loads. Most studies of pollen flow in distylous species have been based on populations in which the two floral morphs occur in a 1:1 ratio, and are randomly distributed in the population. The previously described asymmetric pollen flow patterns are typical of such populations. Because of the clonal nature of Menyanthes, there is a considerable amount of variation in floral morph frequency among populations. Pollen flow patterns from six different populations of Menyanthes were compared in order to investigate the influence of morph POLLEN FLOW / 45 frequency on the composition of stigmatic pollen loads. In four of the six populations studied, the morph frequency deviated significantly from the expected 1:1 ratio. In previous studies of pollen flow in distylous species, populations were sampled on one date during the flowering season. This approach assumes that there is no seasonal variation in pollen flow. Only one study has measured seasonal changes in pollen flow in a distylous species. In Primula vulgaris, stigmatic pollen loads were collected on two dates, one at the beginning of the flowering season, the other at the end (Ornduff 1979). Stigmas collected at the end of the season had twice as many pollen grains deposited on them as those collected at the beginning of the season. The relative proportions of pin and thrum pollen deposited on pin stigmas was the same at the beginning and end of the flowering season, but on thrum stigmas there was a threefold increase in the proportion of pin pollen received between the two collection dates. These results indicate that in populations of Primula vulgaris, the size and composition of stigmatic pollen loads are not constant over the flowering season. Seasonal variation in the size and/or composition of the stigmatic pollen loads of Menyanthes was examined in six populations of Menyanthes by collecting stigmas at regular intervals during the flowering season. B. MATERIALS AND METHODS In the spring and summer of 1985, samples of stigmatic pollen loads were collected from flowers of Menyanthes trifoliata at 2-3 day intervals during P O L L E N F L O W / 46 the flowering season. Stigmas were collected from six study areas, all located within the three populations described in Chapter 2. Three areas were chosen from Beaver Lake and are referred to here as BL2 , BL3 and BL4 (Figure 8). At Pinecrest, two study sites were chosen and are designated as P C I and PC2 (Figure 9). One study area was chosen at Stump Lake and is referred to as S L l (Figure 10). Study areas B L 2 , B L 3 , BL4 , S L l and P C I all represented a subset of a larger population. Study area PC2 included all the inflorescences in a single population (Figure 11). (Extensive clonal growth in Menyanthes made it impossible to determine the number of genets in each population, so the number of inflorescences was counted instead.) At each study site, the frequency of the two floral morphs was determined and deviations from the expected isoplethic (1:1) morph ratio were analyzed using the G-statistic (Sokal and Rohlf 1981). A nearest neighbour count was performed and a 2x2 contingency table used to determine whether the two floral morphs were randomly distributed among one another or segregated into monomorphic clumps (Pielou 1961). Pollen frequencies were calculated from the proportion of pin and thrum (or homostyle) pollen produced in each population, adjusted for the frequency of pin and thrum (or homostyle) inflorscences growing in each study area (Ganders 1979). During each visit to each study area, 40 stigmas per floral morph were collected from intact flowers using jeweller's forceps. Stigmas were collected from flowers that had been exposed to pollinators for 2-3 days. Stigmas of Menyanthes trifoliata flowers usually remain receptive for 3-4 days. The anthers POLLEN FLOW / 47 FIGURE 8. Population of Menyanthes trifoliata at Beaver Lake in Stanley Park, Vancouver, B.C. Stippled area indicates extent of population. Shaded areas indicate study areas BL2, BL3 and BL4 where pollen flow and seed set data were collected. POLLEN FLOW / 48 FIGURE 9. Populations of Menyanthes trifoliata located in the Pinecrest area, 0.9 km north of Daisy Lake canal on B.C. Hwy. 99. Stippled areas indicate extent of populations. Shaded areas indicate study areas PCI, PC2 and PC3. POLLEN FLOW / 49 FIGURE 10. Population of Menyanthes trifoliata at Stump Lake in Alice Lake Provincial Park in southwestern B.C. Stippled area indicates extent of population. Shaded areas indicate study areas SLl and SL2. Study area SL2 was a small pond separated from Stump Lake by forest, and includes all plants in that population. FIGURE 11. Populations of /Menyanthes trifoliata located in small ponds in the Pinecrest area. In this location small ponds were separated from each other by portions of a Sphagnum bog. Stippled area indicates the area in each pond where Menyanthes occurred. Shaded areas indicate study areas PC1 and PC2. Study area PC 1 include a large portion of the plants growing in one pond. Study area PC2 included a l l the plants growing in another pond. O P O L L E N F L O W / 51 are wilted by the third day of anthesis, and by the fourth day the corolla begins to wilt. Those flowers with wilted anthers but fresh corollas were chosen as flowers that had been exposed to pollinators for three days, ensuring that stigmas were collected from flowers at the same stage of maturity. Stigmas were collected from flowers of plants selected by chance in the study area. Pin and thrum (or homostyle) stigmas were collected separately and pooled for each floral morph. Pollen samples from stigmas were obtained by acetolyzing the pooled collections of stigmas and suspending the pollen in a 1:1 mixture of lactophenol and glycerin (Erdtman 1960). Pollen counts were made using a haemocytometer and twenty replicates of each sample were counted (Lloyd 1965). Because the pollen size classes of the two floral morphs for each population overlapped slightly, the point of minimum overlap between the two size classes of each population was chosen as a cutoff point. For pollen from Beaver Lake the cutoff point was 43.0 jum, for Stump Lake 44.0 fim, and for Pinecrest 45.0 um. In order to determine the number of pin and thrum (or homostyle) pollen grains in each sample of pooled stigmas, an image splitting module on a Vickers M22 microscope was set at the cutoff point. The module sheared the image of the pollen grains, producing either double images that overlapped when the pollen grain was larger than the cutoff point, or two discrete images when the pollen grain was smaller than the cutoff point. The number of larger and smaller pollen grains was determined for each sample and these numbers were used to estimate the number of pin and thrum (or homostyle) pollen grains present using the following formulae (Ganders 1976): POLLEN FLOW / 52 T = S - a (L + S) b - a and P = L + S - T T= number of thrum (or homostyle) pollen grains P= number of pin pollen grains L= number of pollen grains larger than the cutoff point S= number of pollen grains smaller than the cutoff point a = frequency of pin pollen grains smaller than the cutoff point in the measured pin pollen size distribution b= frequency of thrum (or homostyle) pollen grains larger than the cutoff point in the measured thrum pollen size distribution Stigmas were also collected from emasculated flowers. Emasculations were performed by removing, with jeweller's forceps, all the anthers from newly opened flowers (Ganders 1974). Only those flowers with undehisced anthers were used. Flowers for emasculation were chosen by chance, and after emasculation were tagged with the date, and their stigmas collected three days later. Stigmas were processed as described above, and the number of pin and thrum (or homostyle) pollen grains determined using the same formulae. Stigmatic pollen loads from both intact and emasculated flowers were counted on a per flower basis. Pollen loads of pin and thrum (or homostyle) stigmas were treated separately for each study area. Because of the slight overlap in pollen sizes between floral morphs, estimates of the number of pin P O L L E N F L O W / 53 and thrum (or homostyle) pollen grains deposited on stigmas may not be exact, but should provide a general indication of the proportion of the two pollen types present. Means were calculated for individual collection dates and for all dates combined. Unless otherwise specified, results refer to averages over the whole flowering season, not to averages of individual collection dates. Mean stigmatic pollen loads were compared to pollen frequencies for each study area in order to determine how the proportion of pin and thrum (or homostyle) pollen on stigmas compared to the proportion of pin and thrum (or homostyle) pollen produced in the study area. Pollen frequencies were used for this comparison instead of morph frequencies, because they represented a more accurate estimate of how much pin and thrum (or homostyle) pollen was available for pollination in each population. C. RESULTS The total number of inflorescences within each study area ranged from 86 (PC2) to 433 (BL3) (Table 5). Based on the results of tests using the G-statistic, only BL3 and P C I had the expected isoplethic (1:1) ratio of floral morphs that is usually found in distylous plant populations (Table 5). The other four study areas were anisoplethic with morph ratios ranging from 7:93% pins:thrums at BL4 , to 91:9% pins:thrums at PC2. Nearest neighbour counts and 2x2 contingency tables showed that BL2 and BL4 were the only two study sites in which pins and thrums were randomly distributed (Table 5). In the remaining four sites, B L 3 , S L l , P C I and PC2, plants were grouped into monomorphic clumps. TABLE 5. Columb1 a. expected Representation of f l o r a l morphs In s ix G values indicate deviat ions from the random d i s t r i b u t i o n of f l o r a l morphs in populations of Menyanthes trifoliata in southwestern B r i t i s h expected 1:1 morph frequency. X' values indicate dev ia t ions from the the population (based on nearest neighbour counts) . Inf lorescences in population F1ora1 Populat ion morph Number Total Percentage G (df=1) P X 1 (df=3) P BL2 p 1 n thrum 205 71 276 74 . 3 25.7 67 . 89 <0.001 O. 32 NS BL3 p in thrum 218 215 433 50.0 50.0 0.03 NS 133.27 <0.001 BL4 p in thrum 14 197 211 6.6 93.4 189.49 <0.001 2 . 25 NS PC 1 pin thrum 100 89 189 52.9 47 . 1 0.53 NS 24.88 <0.001 PC2 p in thrum 78 8 86 90. 7 9.3 63 . 74 <0.001 45 .06 <0.001 SL 1 p in homostyle 212 77 289 73.4 26.6 65.58 <0.001 30.91 <0.001 , o t-1 w O 3 P O L L E N FLOW / 55 Because the number of pollen grains produced per flower varied between floral morphs, the frequencj' of pin and thrum (or pin and homostyle) pollen produced in each population was not exactly the same as the frequency of pin and thrum (or pin and homostyle) inflorescences in each population. Pollen frequencies ranged from 8:92% pimthrum pollen (BL4) to 90:10% pimthrum pollen (PC2) (Table 6). A t sites BL2 , BL3 and PC2 there was significantly more pin pollen than thrum pollen produced in the population. Similarily, at S L l the frequency of pin pollen was significantly greater than that of homostyle pollen. Site BL4 was the only study area in which the frequency of thrum pollen was significantly greater than the frequency of pin pollen. A t P C I , there was no significant difference in pin and thrum pollen frequencies. Approximately 1% of the pollen produced in each population was deposited on conspecific stigmas (Table 7). The proportions of compatible and incompatible pollen deposited on stigmas of flowers of Menyanthes from the six populations studied were very similar to pollen frequencies in the population (Table 8). A t BL2 , BL3 , BL4 and PC2 where there was a significant difference between floral morphs in the percentage of compatible pollen received, this difference corresponded to differences in pollen frequencies in the population (Table 8). A t BL2 , BL3 and PC2, the frequency of pin pollen was significantly greater (p< 0.001) than thrum pollen, and thrum stigmas received a significantly higher (BL2: p<0.01, BL3 p<0.01, PC2 p<0.05) percentage of compatible pollen than pins (Table 8). Similarly, at BL4 the frequency of thrum pollen in the population was significantly greater (p< 0.001), and pins received a significantly higher (p<0.01) percentage of compatible pollen than thrums. At P C I , the frequency of TABLE 6. Frequency of pin, thrum and homostyle pollen produced In six populations of Menyanthes trifoliata. Pollen frequencies were calculated from the production of pin and thrum (or pin and homostyle) pollen adjusted for the frequency of the two f l o r a l morphs in the population. G values Indicate deviations from a 1:1 pollen frequency In the population. Pollen production: N=20 flowers per f l o r a l morph in each population. Popu1 at ion Floral morph Pol 1 en product 1 on per flower Percentage of pol1 en produced in anthers Morph frequency in popu1 at 1 on Pol 1 en frequency in popu1 at 1 on G (df=1) P BL2 pin 28.282 65 . 4 0 .743 0 .845 52.37 <0 .001 thrum 14,980 (676) 34 .6 0 . 257 0. . 155 BL3 p 1 n 16,266 (3452) 66 . 7 0. . 500 0. .667 1 1 .38 <0. .001 thrum 8,898 (2451) 33. . 3 0. .500 0. . 333 BL4 p i n 11, 1 10 53 .4 0. .066 0. .075 42 .67 <0. ,001 thrum 9,693 (2168) 46 .6 0, ,934 0, .925 PC 1 p i n 14,732 (1688) 46. 2 0. 529 0. .491 0.38 NS thrum 17,149 (885) 53 8 0. .471 0. . 509 PC2 p i n 17,153 46. . 2 0. .907 0. 893 70.57 <0. 001 thrum 19,986 53 . 8 0. 093 0, . 107 SL1 p i n 17,783 64 . 0 0. .734 0. 831 46.62 <0 .001 homosty1e 9,990 (1754) 36 . 0 0. 266 0. . 169 o 3 TABLE 7. Percentage of pin and thrum (or homostyle) po l len p a r t i c i p a t i n g In compatible p o l l i n a t i o n s 1n s ix populat ions of Menyanthes trifoliata. Populat ion Percentage p1n pol1 en part 1dpat ing in compatible pol1 inat ions Percentage thrum po11 en p a r t i c i p a t i n g In compatible pol1 inat ions BL2 BL3 0.46 0.97 0. 75 1.12 BL4 PC 1 PC2 SL 1 0. 70 0.58 0.41 0. 1 1 3.22 0.40 0. 30 0. 19* X for a l l study areas comb 1ned * percentage homostyle po l len 0.50 0. 99 Tl O tr1 O 3 TABLE 8. Percentage of po l len deposited on stigmas of p i n , s ix popula t ions . Values are the mean and standard deviat ion c o l l e c t e d at regular in terva ls during the f lowering season, during the whole f lowering season. thrum and homostyle f lowers of Menyanthes trifoliata from ( in parentheses). Means are ca lcu la ted from samples N represents the tota l number of stigmas c o l l e c t e d Percentage Percentage Pol len of p in of thrum Populat ion Stigma frequency N po l len p o l l e n BL2 pi n 0 .845 190 73 .0 (9 .90) 27 . 0 (9.90) thrum 0. . 155 175 61 . 4 (14 .38) 38. 6 ( 14.38) BL3 p 1 n 0 .677 354 69 .6 (19 . 18) 30. 4 (19.18) thrum 0 . 333 333 57 .5 (20 .40) 42 . 5 (20.40) BL4 p 1 n 0. 075 151 32 .7 ( 19. 85) 67. 3 (19.85) thrum 0. 925 320 31 . 1 ( 14. 15) 68 . 9 ( 14.15) PC 1 p 1 n 0. 491 202 69 .9 ( 14. 08) 30. 1 (14.08) thrum 0. 509 205 48 . 4 ( 12. 51) 51 . 6 (12.51) PC2 p 1 n 0. 893 163 81 2 ( 13. 43) 18. 4 ( 13.43) thrum O. 107 43 69 2 (31 . 60) 30. 8 (31.60) SL 1 p i n 0. 831 245 85 . 3 (14. 98) 14.7 '* (14.98) homosty1e 0. 169 160 24 . 0 (30. 30) 76.0* (30.30) Tl o f f w z Tl f O 3 * homostyle po l len oo P O L L E N F L O W / 59 pin and thrum pollen in the population was equal, and there was no significant difference between floral morphs in the percentage of compatible pollen received. There was a very high correlation (r = 0.93, p<0.001) between the percentage of compatible pollen deposited on stigmas and the percentage of compatible pollen produced in each population. As the frequency of thrum pollen produced increased among populations, so did the percentage of thrum pollen deposited on pin stigmas in those populations (Figure 12). The same was true for pin pollen on thrum stigmas. However, this relationship was not proportional. In the five populations where one form of pollen was more abundant than the other, stigmas of the rarer floral morph received a slightly lower percentage of compatible pollen than expected (based on the percentage of compatible pollen produced in the population) (Table 8). In the same populations, stigmas of the more abundant floral morph received a slightly higher percentage of compatible pollen than expected. As was found in other populations of Menyanthes, at SL1 the percentage of compatible pollen deposited on pin stigmas (14.7%, s.d= 14.98) was very similar to the percentage produced in the population (0.169). Homostyles from SL1 were the only flowers in which the composition of stigmatic pollen loads was markedly different from the frequency of pollen produced in the population (Figure 12, Table 8). The frequency of pin pollen produced at SL1 (0.831) was much greater than that of homostyle pollen (0.169); on this basis homostjde stigmas should be expected to have received an excess of compatible pollen. However, homostyle stigmas received an average of only 24.0% (s.d. = 30.30) POLLEN FLOW / 60 FIGURE 12. The percentage of compatible pollen deposited on stigmas of pins, thrums and homostyles in six populations of Menyanthes trifoliata plotted against the percentage of compatible pollen produced in each population. Each point represents one floral morph from each of the six populations: BL2, BL3, BL4, PCI, PC2 and SLl. Closed circles represent pins, open circles represent thrums and the triangle represents homostyles. The line is the least squares regression for pins and thrums (homostyle was excluded). Correlation: r = 0.93, p<0.0001. -j——i 1 1 1 1 1 r 20 4 0 6 0 8 0 % Compat ib le Po l len Produced in Popula t ion P O L L E N F L O W / 61 compatible pollen. Study area PC2 consisted of a pin/thrum population structured similarly to the pin/homostyle population at S L l . A t PC2, the frequency of pin pollen (0.893) was much greater than the frequency of thrum pollen (0.107), and the distribution of floral morphs in the population was non-random. The proportion of compatible pollen deposited on pin stigmas at PC2 (18.4%, s.d= 13.43) corresponded well with the frequency of compatible pollen in the population (0.107), and was very similar to that found for pin stigmas at S L l . Similarly, the proportion of compatible pollen deposited on thrum stigmas from PC2 (69.2%, s.d = 31.60) corresponded well with the frequency of compatible pollen produced in the population (0.893), and was much greater than that received by pin stigmas. However, there was a remarkable difference between the proportion of compatible pollen received by thrums from PC2 and homostyles from S L l ; the thrums received 45.2 percentage points more compatible pollen than the homosytles. Although there was a very high correlation between the percentage of compatible pollen deposited on stigmas of Menyanthes and pollen frequency in the population, the correlation between the numbers of compatible and incompatible pollen grains received and pollen frequency was not as high (r = 0.69, p<0.05). In all study areas except P C I , the frequency of one pollen form was significantly greater than the frequency of the other (Table 6), yet sites BL3 and BL4 were the only study areas in which there was a significant difference between floral morphs in the number of compatible pollen grains received. At BL3 , where pin pollen frequency was higher (0.677), thrums received a P O L L E N FLOW / 62 significantly higher (p<0.05) number of compatible pollen grains than pins and pins received a significantly higher (p<0.05) number of incompatible pollen grains than thrums (Table 9). A t BL4 , where thrum pollen was more abundant (0.925), pins received a significantly higher (p<0.05) number of compatible pollen grains than thrums, but there was no significant difference between floral morphs in the number of incompatible pollen grains received (Table 9). A t P C I , where pin and thrum pollen frequencies were equal, there was no significant difference between floral morphs in the number of compatible or incompatible pollen grains received. In the remaining three study areas (BL2, PC2, S L l ) , pin pollen was much more abundant than the other pollen form, yet there was no significant difference between floral morphs (within each population) in the number of compatible pollen grains received. In addition, at PC2 and S L l there was no significant difference between floral morphs in the number of incompatible pollen grains received. However, at BL2 pins received a significantly higher (p<0.05) number of incompatible pollen grains than thrums. At BL2 , the lack of correlation between the number of compatible pollen grains received and pollen frequency was probably a result of differences in the size of pollen loads between the two floral morphs. At BL2 , the frequency of pin pollen in the population was 0.845, and thrum stigmas received a significantly higher percentage of compatible pollen (69.7%) than pins (27.0%). But, pin pollen loads were almost twice the size of thrum pollen loads, consequently the actual number of compatible pollen grains received by the two floral morphs was similar. Even though thrums at PC2 received at least four times as much compatible pollen as pins, this difference was not statistically significant, perhaps as a result of small sample sizes. TABLE 9. Composition of st igmat ic po l len loads of p i n , thrum and homostyle flowers of Menyanthes trifoliata from s i x popula t ions . Values are the mean and standard deviat ion (1n parentheses). Means are ca lcu la ted from samples c o l l e c t e d at regular Intervals during the f lowering season. N represents the total number of stigmas c o l l e c t e d during the whole f lowering season. Number of Number of Total pin po l len thrum po l len number of Populat ion Stigma N grains grains po l len grains BL2 BL3 BL4 PC 1 PC2 SL 1 p i n thrum p 1 n thrum p i n thrum pin thrum p 1 n thrum pi n homostyle 190 175 354 333 151 320 202 205 163 43 245 160 290.5 (143.25) 113.1 (76.51) 403.6 130.2 (85.96) 86.7 (67.50) 216.9 257.8 (130.77) 99.9 (57.54) 357.7 172.6 (89.70) 129.7 (95.83) 302.3 313.5 (325.83) 466.O 154.6 (75.53) 232.3 59.6 (54.25) 195.9 75.3 (45.28) 145.8 118.9 (101.10) 38.9 (59.98) 157.8 168.1 (174.52) 74.8 (23.46) 242.9 204.9 (46.56) 39.9* (44.40) 244.9 104.3 (68.17) 329.5* (228.23) 433.8 152.5 (99.05) 77.7 (53.94) 136.3 (108.93) 70.5 (70.03) 192.63) 126.56) 128.81 ) 1 16.74) 410.34) 119.85) 159.75) 1 14.03) 154.03) 197.90) 57.65) 192.20) homostyle p o l l e n O 3 OS CO P O L L E N F L O W / 64 Comparison of the actual number of compatible pollen grains deposited on stigmas of PC2 thrums and SL1 homostyles revealed that the thrums received an average of 1.6 times more compatible pollen grains than the homostyles. However, the number of incompatible pollen grains deposited on stigmas of SL1 homostyles was 4.4 times greater than the number received by PC2 thrums. (Pins from the two populations received a nearly identical number of compatible pollen grains (PC2: 38.9, s.d. = 59.98; SL1 : 39.9, s.d. = 44.40).) Patterns of pollen deposition on stigmas over the flowering season were very erratic (Figures 13-18). There was a great deal of variation in the size and composition of stigmatic pollen loads during the flowering season, and there were few patterns in common among the six study areas. Sites P C I and PC2 were the only two populationsn that showed similar pollen deposition patterns in the two floral morphs during the flowering season (Figures 16 and 17). For pins and thrums from BL3 , BL4 , P C I and PC2, and homostyles from SL1 the total number of pollen grains deposited on stigmas seemed to decrease towards the end of the season. However, for pins and thrums from BL2 total stigmatic pollen loads increased in size towards the end of the flowering season. For pins at S L 1 , the total number of pollen grains received was similar at the beginning and at the end of the season. Results were highly variable from the four stud}' areas in which emasculations were performed. There was considerable variation in the composition of stigmatic pollen loads among the different floral morphs on different- dates, and in different populations (Tables 10-13). At BL2 , the composition of stigmatic POLLEN FLOW / 65 FIGURE 13. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL2 during the flowering season of 1985. For each collection date, the number of pin and thrum pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of thrum pollen grains deposited on stigmas. Pi n S t i g m a s 6 0 0 -5 0 0 -to 4 0 0 -< GM 3 0 0 -LO 2 0 0 -z O 1 0 0 -Q LLI 0 -1-U) o o_ IU a z < rr 6 0 0 -O z 5 0 0 -LLI _l _l 4 0 0 -o Q_ 3 0 0 -LL o cr 2 0 0 -LU m 1 0 0 -ZD Z 0 -i i 1 1 r 5 /16 19 22 2 4 2 8 D A T E T h r u m S t i g m a s i 1 1 r 5 /16 19 22 24 28 D A T E POLLEN FLOW / 66 FIGURE 14. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL3 during the flowering season of 1985. For each collection date, the number of pin and thrum pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of thrum pollen grains deposited on stigmas. Pin S t i gmas i f ) < o «-to z O Q LU f-Ln O CL LU Q CO Z < rr CD z LU O o_ LL O ce LU CO ZD T 1 1 1 r 5 /19 2 2 2 4 2 8 6/1 4 D A T E Thrum St igmas i 1 1 r 6 1 2 1 5 1 8 1 1 1 1 1 r— 5/19 2 2 2 4 2 8 6/1 4 D A T E T 1 1 1 8 1 2 1 5 1 8 POLLEN FLOW / 67 FIGURE 15. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population BL4 during the flowering season of 1985. For each collection date, the number of pin and thrum pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of thrum pollen grains deposited on stigmas. P i n S t i g m a s 5 / 2 2 T r 24 28 6/1 < cr CD LU 4 8 D A T E -i 1 1 r 12 15 18 21 T h r u m S t i g m a s O 4 0 0 CL LL 3 0 0 H O cr 2 0 0 -| LU | 1 0 0 -z o 1 1 1 1 1 1 1 r 5 /22 24 28 6/1 4 8 12 15 18 21 D A T E POLLEN FLOW / 68 FIGURE 16. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population PCI during the flowering season of 1985. For each collection date, the number of pin and thrum pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of thrum pollen grains deposited on stigmas. Pin S t i gmas CO Z o o LU h-00 O Q_ LU Q 00 5 0 0 -I 2 4 0 0 -| O 3 0 0 H 2 0 0 -1 0 0 -O 6 / 3 < or O LU _ l _ l O CL LL o cc LU CO ZD Z Thrum S t i gmas 6 / 3 8 T 1 r 11 15 18 21 D A T E POLLEN FLOW / 69 FIGURE 17. The number of pollen grains deposited on stigmas of pin and thrum flowers of Menyanthes trifoliata in population PC2 during the flowering season of 1985. For each collection date, the number of pin and thrum pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of thrum pollen grains deposited on stigmas. P i n S t i g m a s to < o 10 z O Q LU h-00 O CL LU Q CO z < cr O 6 / 3 8 11 1 5 1 8 2 1 D A T E Thrum S t i g m a s 6 / 3 8 11 D A T E POLLEN FLOW / 70 FIGURE 18. The number of pollen grains deposited on stigmas of pin and homostyle flowers of Menyanthes trifoliata in population SLl during the flowering season of 1985. For each collection date, the number of pin and homostyle pollen grains deposited on stigmas was determined from pooled samples of approximately 40 stigmas per floral morph. Closed circles and solid lines represent the number of pin pollen grains deposited on stigmas. Open circles and broken lines represent the number of homostyle pollen grains deposited on stigmas. ^ P in St igmas < I O 4 0 0 -5 / 1 8 21 24 26 31 D A T E TABLE 10. Expected and observed po l len frequencies on stigmas of p in and thrum flowers of Menyanthes trifoliata populat ion BL2. N=10 stigmas per f l o r a l morph for each c o l l e c t i o n date. Expected values are based on random p o l l i n a t i o n and are ca lcu la ted from the frequency of pol len produced in the populat ion. X J values indicate the dev ia t ion of observed po l len gra in deposi t ion from random p o l l i n a t i o n . P o l l i n a t i o n regime: R= random p o l l i n a t i o n , s i g n i f i c a n t assor ta t l ve p o l l i n a t i o n . Pol 1 en type F lo ra l P o l l i n a t i o n morph Date Pin Thrum X ! (df=1) regime P1n 5/16 Obs . Exp. 146 .4 140.6 20.0 25.9 1 .57 5/19 Obs . Exp. 164 .0 161 .7 27 .9 29.7 0.14 thrum 5/ 16 Obs . Exp. 93 . 2 253 .4 206 . 7 46.5 653 . 2 5/19 Obs . Exp. 143.5 151 .O 35 . 2 27.7 2.4 TABLE 11. Expected and observed po l len frequencies on stigmas of pin and thrum flowers of Menyanthes trifoliata from populat ion BL3. N=10 stigmas per f l o r a l morph for each c o l l e c t i o n date. Expected values are based on random p o l l i n a t i o n and are ca lcu la ted from the frequency of pol len produced in the populat ion. X' values indicate the dev ia t ion of observed p o l l e n gra in deposi t ion from random p o l l i n a t i o n . P o l l i n a t i o n regime: R= random p o l l i n a t i o n , D= s i g n i f i c a n t d i s a s s o r t a t i v e p o l l i n a t i o n , A= s i g n i f i c a n t assor ta t lve p o l l i n a t i o n . Ca lcu la t ions of percentage e f f i c i e n c y of d i s s a s o r t a t i v e p o l l i n a t i o n explained in text. Pol 1 en type F1ora1 morph Date P1n Thrum X' (df=1) Pol 1(nation regime Percentage e f f i c i e n c y of d i s a s s o r t a t i v e pol1inat ion Pin 5/16 Obs . Exp. 202 . 2 157.3 32.6 77 .5 38.9 5/19 Obs . Exp. 357 .6 248. 1 12.7 122 . 1 146 . 3 5/22 Obs . Exp. 37 . 7 34.8 14.2 17.1 0.98 5/24 Obs . Exp. 40.7 90.9 94 .9 44.7 84. 1 55 . 2 Thrum 5/16 5/19 5/22 5/24 Obs . Exp. Obs . Exp. Obs. Exp. Obs . Exp. 179. 1 202.6 366.7 276.8 58.8 53.6 6.9 16.7 123 . 3 99.8 46.5 136.4 21.1 26.4 18.0 8.2 8.2 88.5 1 .5 12.5 65.9 O f f w f O 3 t o TABLE 12. Expected and observed po l len frequencies on stigmas of p in and thrum flowers of Menyanthes trifoilata from populat ion PC 1 . N=10 stigmas per f l o r a l morph for each c o l l e c t i o n date. Expected values are based on random p o l l i n a t i o n and are ca lcu la ted from the frequency of po l len produced in the populat ion. X' values Indicate the dev ia t ion of observed po l len gra in deposi t ion from random p o l l i n a t i o n . P o l l i n a t i o n regime: R= random p o l l i n a t i o n , D= s i g n i f i c a n t d isassor ta t 1ve p o l l i n a t i o n , A= s i g n i f i c a n t assor ta t lve p o l l i n a t i o n . Ca lcu la t ions of percentage e f f i c i e n c y of d i s s a s o r t a t i v e p o l l i n a t i o n explained in text. Pol 1 en type F1 oral morph Date P in Thrum X 1 (df=1) Pol 1inat1on regime Percentage e f f i c i e n c y of di sassortat1ve pol11 nat ion Pin 6/8 Obs . Exp. 353.9 315.0 287 . 7 326 .6 9.4 6/11 Obs . Exp. 437 . 7 378 . 1 332.3 391 .9 18 . 5 6/15 Obs . Exp. 118.3 124.8 135.9 129.4 0.65 6/18 Obs. Exp. 121.4 149.3 182.7 154.8 10.2 18.7 Thrum 6/8 6/11 6/15 6/18 Obs . Exp. Obs . Exp. Obs. Exp. Obs. Exp. 255.6 212.8 173 .8 148 . 1 189.8 150.0 56.7 39.3 177 .0 220.6 127.9 153.6 1 15.7 155 .6 23.3 40. 7 16 .9 8.8 20.8 15. 1 19.4 16.7 25.6 42.8 TJ O f f z Tl f O 3 CO TABLE 13. Expected and observed po l len frequencies on stigmas of p in and homostyle flowers of Menyanthes trifoliata from populat ion SL1. N=10 stigmas per f l o r a l morph for each c o l l e c t i o n date. Expected values are based on random p o l l i n a t i o n and are ca lcua l ted from the frequency of pol len produced in the populat ion. X' values indicate the dev ia t ion of observed po l len gra in deposi t ion from random p o l l i n a t i o n . P o l l i n a t i o n regime: R= random p o l l i n a t i o n , D= s i g n i f i c a n t d i s a s s o r t a t i v e p o l l i n a t i o n , A= s i g n i f i c a n t assor ta t ive p o l l i n a t i o n . Ca lcu la t ions of percentage e f f i c i e n c y of d issasor ta t 1ve p o l l i n a t i o n explained in text. Pol 1 en type F lo ra l morph Date P i n Homosty1e (df=1 ) Pol 1inat ion regime Percentage e f f i c i e n c y of d i s a s s o r t a t i v e pol1inat ion Pin 5/18 Obs. 2.0 1.0 0.35 R Exp. 2.5 2.5 5/21 Obs. 52.4 10.5 0.002 R Exp. 52.2 10.G 5/24 Obs. 40.G 30.0 33.0 0 30.8 Exp. 58.7 11.9 Homostyle 5/18 Obs. 49.9 52.8 8G.7 A Exp. 85.3 17.4 HJ O 5/21 Obs. 9.1 24.8 76.9 A tT1 Exp. 28.2 5.7 2 5/24 Obs. 161.8 79.2 43.8 A f Exp. 200.3 40.7 O 3 4^ P O L L E N F L O W / 75 loads of pins was not significantly different than that expected from random pollination (Table 10). The same was true for thrums at BL2 on one date, but on the other date they received a significant amount of assortative pollination. Pin stigmas from BL3 were randomly pollinated on one date, received a significant amount of assortative pollination on two dates, and were disassortatively pollinated on the fourth (Table 11). Thrums stigmas from BL3 received a significant amount of assortative pollination on three dates, and were randomly pollinated on the fourth. At P C I , pin stigmas experienced assortative, random and disassortaive pollinations on different dates (Table 12). However, thrum stigmas received a significant amount of disassortative pollination on all dates sampled. At SL1 , pin stigmas were either randomly or disassortatively pollinated, whereas homostyles received a significant amount of assortative pollination on all dates measured (Table 13). The formula E = (o - r)/(d - r) measures the efficiency of distyly in promoting disassortative pollination in emasculated flowers, where E is the disassortative component of the pollen load, o is the observed frequency of one of the pollen forms, r is the expected frequency of that pollen form with random pollination, and d is the expected frequency of that pollen form with disassortative pollination (Ganders 1974). Values of E range from 0% (random pollination) to 100% (complete disassortative pollination), and were calculated for the four populations of Menyanthes in which significant levels of disassortative pollination were detected (BL2, BL3 , P C I and S L l ) (Tables 10-13). There was a considerable amount of variation in the degree of disassortative pollination experienced on different dates throughout the flowering season. Values ranged P O L L E N F L O W / 76 from 16.7% to 65.9%, and averaged 34.9% (s.d= 18.59) for pins and 34.1% (s.d = 20.48) for thrums on the occasions when the flowers were disassortatively pollinated (homostyles were not disassortatively pollinated). D. DISCUSSION None of the six populations of Menyanthes studied had the typical population structure of a distylous plant: a 1:1 ratio of floral morphs randomly distributed among one another. Price and Barrett (1984) have found that in Pontederia cordata, a combination of clonal reproduction and pollinator behavior resulted in a 75% chance that plants growing next to each other would be of the same form. While little is known of pollinator activity in populations of Menyanthes, the clonal nature of the species is a major factor influencing the population structure. Hewett (1964) concluded that vegetative propagation was the most important method of reproduction in Menyanthes, and he stated that seedlings had not been reported from nature. I did find seedlings of Menyanthes germinating in the mud in the ponds at Pinecrest (Figure 19), but it appeared that vegetative propagation was a much more common form of reproduction. The low percentage of pollen (<1%) participating in legitimate pollinations in populations of Menyanthes is similar to that found in other distylous species (Onrduff 1970a, 1970b, 1971, 1976, 1979, 1980a; Ganders 1974; Weller 1980; Philipp and Schou 1981; Rama Swamy and Bir Bahadur 1984; Nicholls 1985), and seems to be a common feature of distylous breeding systems. Since more than 60% of the pollen produced in populations of Menyanthes was removed from P O L L E N F L O W / 77 F IGURE 19. Seedlings of Menyanthes trifoliata germinating in the mud near ponds in the Pinecrest area of southwestern British Columbia. P O L L E N F L O W / 78 anthers by pollinators (Table 3), the low percentage of Menyanthes pollen reaching conspecific stigmas was not the result of a lack of pollinator activity. The bulk of the pollen produced in populations of Menyanthes must have been lost through vagaries of the pollination process. In flowers of Menyanthes trifoliata, a considerable amount of variation existed among populations, between floral morphs and among individual collection dates in the size and composition of stigmatic pollen loads. Pollen deposition patterns fluctuated erratically during the flowering season, and were very different among the six populations studied. Within each study area, flowers from both floral morphs experienced either random, assortative or disassortative pollination depending on the date the samples were collected. The magnitude of disassorative pollination varied from 16.7% to 65.9%. Thrums from P C I were the only flowers that received a significant amount of disassortative pollination on all dates measured, while homostyles from SL1 received a significant amount of assortative pollinatation on all dates measured. Glover and Barrett (1986) found similar intermorph and interpopulational variation in the composition of stigmatic pollen loads of emasculated flowers of tristylous Pontederia cordata. The long-styled floral morph was disassortatively pollinated in all four populations examined, but the type of pollination regime experienced by the mid-styled and short-styled morph varied among populations. The mid-styled morph experienced both disassortative and random pollination and the short-styled morph experienced all three pollination regimes: random, disassortative and assortative pollination. Among those flowers that were P O L L E N F L O W / 79 disassortatively pollinated, the degree of disassortative pollination ranged 7.1% to 39.3%. Individual variation in the composition of stigmatic pollen loads has also been found in populations of distylous Hedyotis caerulea (Ornduff 1980b) and Linum tenuifolium (Nicholls 1985). Ornduff (1980b) found that in H. caerulea up to 21% of the stigmas held no pollen at all. Results from these studies of pollen flow in various heterostylous species (including Menyanthes trifoliata), indicate that pollen flow is not as uniform among individuals or during the flowering season as was once assumed. The size and composition of stigmatic pollen loads appears to be highly variable among individuals of each floral morph, between floral morphs, among different days during the flowering season and among populations. These results suggest that erratic pollen deposition may be. a common feature of pollen flow in distylous plant populations. Despite the apparent erratic nature of pollen flow in populations of Menyanthes trifoliata, some general trends were evident. Among the six populations of Menyanthes studied, there was a high correlation between the frequency of pollen produced in the population and the composition of stigmatic pollen loads. Richards and Ibrahim (1978) found a similar relationship between the proportion of pin and thrum pollen deposited on stigmas and the frequency of floral morphs in populations of Primula veris. However, Ornduff (1976, 1978, 1980b, 1986) was unable to find a similar relationship in the various distylous species he studied. P O L L E N F L O W / 80 In populations of Menyanthes, the high correlation between pollen frequency in the population and the composition of stigmatic pollen loads indicates that in general, pollen flow was close to what would be expected with random pollination. However, this relationship was not proportional. In anisoplethic populations of Menyanthes, stigmas of the more abundant floral morph received a slightly higher percentage of compatible pollen than expected based on random pollination. Whereas, stigmas of the rarer floral morph received a slightly lower percentage of compatible pollen than expected. This is significant because it suggests that morphological distyly is successful in compensating for the rarity of one floral form by increasing the proportion of compatible pollen received by the opposite floral form. Distyly seems to buffer the effects of drastic differences in morph or pollen frequency in the population. Comparison of stigmatic pollen loads of the homostyles from S L l to pollen loads of thrums from PC2 revealed a remarkable difference in the proportion of compatible pollen received. Thrums from PC2 received approximately 45 percentage points more compatible pollen than the homostyles from SL2. But, when the actual number of compatible pollen grains received by the thrums and homostyles was compared, the difference was not as great. Thrums received approximately 1.6 times more compatible pollen grains than homostyles. The major difference in stigmatic pollen loads between the thrums and homostyles was in the number of incompatible pollen grains received. Homostyles received more than four times the number of incompatible pollen grains received by thrums. POLLEN FLOW / 81 Except for the difference in style length, the thrums and homostyles were morphologically identical. They both grew in similarly structured populations. The composition of pin stigmatic pollen loads was very similar in both populations indicating that pollen flow, patterns were similar in the two populations. Therefore, differences in the composition of stigmatic pollen loads between the thrums and homostyles can be attributed to the difference in style length found in the two floral morphs. Because stigmas and anthers in homostyle flowers are close to each other, they received a much higher number of own-form pollen grains than did thrums. In addition, the reciprocal stamen/style dimorphism in the pin/thrum population resulted in thrum stigmas receiving a higher number of compatible pollen grains than homostyle stigmas. The separation of stigmas and anthers in pin and thrum flowers in Menyanthes reduces the number of incompatible pollen grains received, and the reciprocal placement of stigmas and anthers does appear to increase the number of compatible pollen grains received. These results provide quantitative evidence in support of Darwin's (1877) hypothesis that morphological distyly increases the amount of compatible pollen deposited on stigmas of diallelic self-incompatible plants. V. SEED SET A. INTRODUCTION With the diallelic self-incompatibility mechanism found in distylous plants, the chances of a stigma receiving incompatible pollen are great. If one assumes that seed set is limited by a lack of compatible pollinations, then any mechanism that would increase the amount of compatible pollen received should increase the fecundity of the plant. Darwin (1877) suggested that because of the reciprocal placement of stigmas and anthers, flowers of distylous plants should receive a greater amount of compatible pollen than flowers of a diallelic plant lacking the floral dimorphism. Ganders (1974) hypothesized that morphological distyly might result in an increase in fecundity of a diallelic self-incompatible plant by increasing the number of compatible pollen grains received. An unusual population of Mitchella repens consisting of pins, thrums and self-incompatible homostyles allowed Ganders (1975a) to test this hypothesis. Except for the length of their styles, the homostyle flowers were identical to the thrum flowers. Ganders suggested that any difference in seed set that existed between thrums and homostyles could be attributed to the influence of the stamen/style dimorphism on the number of compatible pollen grains received. Ganders' results show that the thrums had significantly higher seed set than the homostyles and he suggested that this indicated that in diallelic self-incompatible plants, "distyly may be associated with a greater reproductive fitness than homostyly". 82 SEED SET / 83 Although Ganders (1975a) was able to demonstrate that the distylous flowers were more fecund than the homostylous flowers, he lacked the pollen flow data to confirm that differences in seed set between the two floral morphs resulted from differences in the number of compatible pollen grains received. Many factors have been shown to influence seed set in self-incompatible plants: the quantity and efficiency of pollinators, the density, spatial pattern and number of mating groups in the population, environmental conditions during pollination, fertilization and embryo development, resource availability for fruit and seed maturation, and pollen availability (Shore and Barrett 1984). Without pollen flow data, it is impossible to determine whether the observed difference in seed set between pins and homostyles of Mitchella repens resulted from pollen limitation or from some other factor. I investigated the relationship between the number of compatible pollen grains received and seed set in six populations of Menyanthes trifoliata in order to determine whether pollen availability was a major factor influencing seed set. For each floral morph . in each study area, the number of compatible pollen grains received (Chapter 4) was compared to the number of seeds set. Results from pollination intensity experiments (Chapter 3) showed that an average of six compatible pollen grains are required to fertilize each ovule in flowers of Menyanthes. Using this number as a basis for comparison, it was possible to determine whether observed seed set in Menyanthes corresponded to the number of compatible pollen grains received. In one population, an attempt was made to increase seed set by increasing the number of compatible pollen grains deposited on stigmas. Flowers were hand-pollinated with large amounts of compatible pollen, SEED SET / 84 and then the resultant seed set was compared to seed set in naturally pollinated flowers in the same population. In addition, seed set of homostyles in a pin/homostyle population of Menyanthes was compared to seed set of thrums in a pin/thrum population. Like the homostyles in the population of Mitchella repens (Ganders 1975a), the homostylous flowers of Menyanthes are morphologically identical to thrum flowers except for the length of their styles. In addition, evidence (below) indicates that homostyles of Menyanthes are self-incompatible and equivalent to thrums in incompatibility behavior. Examination of stigmatic pollen loads revealed that Menyanthes thrums received 1.6 times more compatible pollen grains than the homostyles (Chapter 4), and I attribute this difference to the presence of the stamen/style dimorphism in flowers from the pin/thrum population. If, as Ganders (1974) suggests, distyly is associated with a greater reproductive fitness than homostjdy, then the higher number of compatible pollen grains observed on thrum stigmas should be associated with higher seed set. B. MATERIALS AND METHODS I collected seed set data in the same six populations of Menyanthes in which pollen flow data were collected (BL2, BL3 , BL4, P C I , PC2 and S L l ) . Seed set data were collected from flowers tagged on the same days that stigmas were collected for pollen flow analysis; this gave an estimate of seed set for flowers pollinated during the same period as those from which stigmas had been collected. Ten to twenty flowers per floral morph, at the same stage of anthesis SEED SET / 85 as those from which stigmas were collected, were tagged with the date. Some flowers were selected from the same inflorescence from which stigmas had been collected, and others were on separate inflorescences. Inflorescences with tagged flowers were flagged with fluorescent surveyors' tape for easy relocation. The fruits of the tagged flowers were allowed to mature for 4-5 weeks and collected prior to dehiscence. Percentage seed set was determined by counting the number of seeds and the number of aborted ovules in each fruit. Approximately 60 fruits per floral morph were collected for each population during the course of the flowering season. Bulk samples of undehisced fruits were collected from nine populations. Approximately 300 fruits per floral morph were collected from BL2 , BL3 , BL4 , P C I , PC2 and SL1 . In addition, bulk fruit samples were collected from three monomorphic populations: one from the Stump Lake area (SL2), one from the Pinecrest area (PC3) and one at Strachan Meadow (SMI) (Figures 9, 10, 20). Populations S M I and PC3 were both monomorphic for thrums, while SL2 was monomorphic for homostyles. Up to 3255 fruits were collected from each of the three monomorphic populations. In an attempt to increase seed set, large amounts of compatible pollen were applied by hand to receptive stigmas of flowers growing in the SL1 study area. Recently dehisced anthers from flowers of the opposite form were rubbed on stigmas until all pollen had been deposited (Weller 1980). A total of 14 pin flowers and 14 homostyle flowers were hand-pollinated. These flowers were tagged with the date, their fruits collected 4-5 weeks later and percentage seed set was SEED SET / 86 FIGURE 20. Populations of Menyanthes trifoliata at Strachan Meadow in Cypress Provincial Park in West Vancouver, B.C. Populations were found in a series of small ponds scattered throughout the meadow and along the edges of Yew Lake. Stippled area indicates the location and extent of populations. Shaded areas indicate study area SMI where seed set data were collected. S T R A C H A N M E A D O W L E G E N D H M. t r i f o l i a t a | S t u d y a r e a - S M 1 A N 5 0 m SEED SET / 87 determined. Al l calculations of seed set and percentage seed set were determined on a per fruit basis. Averages of seed set and percentage seed set were calculated for pins and thrums (or homostyles) in each population, and wherever possible for specific collections dates during the season. Unless otherwise specified, results refer to averages for the whole flowering season rather than individual collection dates. C. R E S U L T S In the six populations of Menyanthes studied, there was a very high correlation (r = 0.88, p<0.001) between the average number of compatible pollen grains deposited on stigmas of pins, thrums and homostyles- and the average number of seeds set per capsule (Figure 21). On average, pins from BL4 received the highest number of compatible pollen grains of any floral morph from the six populations studied and had the highest seed set (Table 14). Pins from PCI and PC2 received the lowest number of compatible pollen grains and had the lowest seed set. At BL2 , BL3 , and P C I no significant difference appeared between pins and thrums in the number of seeds set. Similarly, at SL1 no significant difference existed between pins and homostyles in the number of seeds set. However, at BL4 and PC2 the average number of seeds set differed significantly between floral morphs. A t BL4 , pin seed set was significantly higher than thrum SEED SET / 88 FIGURE 21. The number of seeds set per capsule in pins, thrums and homostyles of Menyanthes trifoliata plotted against the number of compatible pollen grains deposited on stigmas. Each point represents one floral morph from each of the six populations studied: BL2, BL3, BL4, PCl, PC2 and SL1. Closed circles represent pins, open circles represent thrums and the triangle represents homostyles. The line is the least squares regression of seed set on the number of compatible pollen grains received. Correlation: r = 0.88, p<0.001. Number of Compatible Pollen Gra ins Deposited on Stigmas TABLE 14. Seed set and percentage seed set in p ins , thrums and homostyles of Menyanthes trifoliata from s ix popula t ions . Values are the mean and standard deviat ion ( in parentheses). Means are ca lcu la ted from samples c o l l e c t e d at regular in te rva ls during the f lowering season. r values represent the c o r r e l a t i o n between the number of compatible p o l l e n grains deposited on stigmas and the number of seeds set for indiv idual c o l l e c t i o n dates dur ing the f lower ing season (see Figures 22-27 for data) . Populat ion F lora l morph Number of compat ible pol1 en grains per stigma Seed set per capsule Percentage seed set BL2 p in thrum 113.1 130.2 30 35 7.0 8 . 3 (5.04) (4.43) 33.5 34 . 2 (30.15) (26.31 ) -0.97 -0.55 BL3 p in thrum 99.9 172.6 71 72 6.4 (3.97) 10.3 (7.32) 25.2 30. 6 (27.84) (27.32) -0. 34 0. 48 BL4 p 1 n thrum 313.5 77.7 31 108 18.8 (4.07) 9.5 (4.79) 71.1 48.8 (24.05) (38.86) -0.09 0.45 PC 1 p in thrum 59 .6 70.5 74 106 3 O 8.0 (2.13) (5. 19) 10. 1 28 . 1 (20.58) (32.09) -0.67 0. 22 PC2 SL1 p in thrum p i n homosty1e 38.9 168 . 1 39.9 104 . 3 56 25 86 61 3.9 (2.76) 16.0 (7.07) 5.8 5. 1 (2.12) (2.79) 7.1 (18.88) 45.7 (29.86) 34. 1 31 .8 (38.23) (31.23) -0.77 -0.60 0.87 -0.87 CO H w o w H oo CO SEED SET / 90 seed set (p< 0.001) and this difference corresponded to a significant difference (p<0.05) between floral morphs in the number of compatible pollen grains received (Table 14). At PC2, thrum seed set was significantly higher (p<0.05) and thrums received at least four times more compatible pollen grains than pins. (The difference in the number of compatible pollen grains received by the two floral morphs was not significant, probably as a result of small sample sizes.) At BL2, PCI and SLl where there was no significant difference between floral morphs in seed set, neither was the number of compatible pollen grains received by the two floral morphs significantly different. At BL3, seed set was similar for the two floral morphs even though thnums recieved significantly more (p<0.05) compatible pollen than pins. Comparison of seed set in the pin/homostyle population to that in the similarly structured pin/thrum population at PC2, revealed that thrums were more fecund than homostyles (Table 14). Thrum seed set at PC2 averaged 16.0 seeds per capsule (s.d. = 7.07), and was significantly greater (p<0.05) than seed set in the homostyles from SLl which averaged 5.1 seeds per capsule (s.d. = 2.79). In general, patterns of seasonal seed set in the six populations of Menyanthes examined were strikingly erratic. Within each study area, seed set fluctuated considerably between collection dates. However, among populations and floral morphs there appeared to be a general trend towards lower seed set towards the end of the flowering season. Although the correlation between seasonal averages of seed set and the number of compatible pollen grains received was very high, within each population, there was little or no correlation between the number of seeds set and the number of compatible pollen grains SEED SET / 91 received on specific dates during the flowering season (Table 14, Figures 22-27). The average number of compatible pollen grains received on each collection date was compared to seed set averages calculated (1) with tagged fruits that failed to set seed (heavy solid line), and (2) without failed fruits (lighter solid line), but neither corresponded very well to the seasonal pollen deposition pattern. The basic difference that appeared between the two sets of seed set calculations was one of magnitude. Pins from S L l were the only plants in which there was a significant positive correlation (r = 0.87, p<0.05) between seasonal seed set patterns and the deposition of compatible pollen (Table 14, Figure 27). Calculation of seasonal averages of percentage seed set for pins, thrums and homostyles of Menyanthes revealed that in the six populations examined, seed set was always below the potential maximum (Table 14). Pins from BL4 had the highest percentage seed set averaging 71.1% (s.d. = 24.05). However, percentage seed set for all other plants in the six populations studied was considerably lower averaging 29.9% (s.d. = 12.58). At BL4, P C I and PC2, there was a significant difference between pins and thrums in the percentage of seeds set. A t BL4 , percentage seed set was significantly higher (p<0.01) in pins than in thrums. At P C I and PC2, percentage seed set was significantly higher in thrums (PCI: p<0.001, PC2: p<0.001). A t BL2 , BL3 and S L l there was no significant difference between floral morphs in percentage seed set. Results from pollination intensity experiments (Chapter 3) showed that in flowers of Menyanthes an average of six compatible pollen grains are required to fertilize each ovule. Comparison of the average number of seeds set on specific SEED SET / 92 FIGURE 22. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL2 during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. m E - 4 0 0 Pin F lowers tn c i_ 0 c o o_ i 1 r 5/16 19 22 24 28 D A T E Thrum F lowers T 1 1 1 r 5/16 19 22 24 2 8 D A T E SEED SET / 93 FIGURE 23. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL3 during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. fx E O) tn O o. a D in c a> i_ O c o o_ 1 0 0 H 0 Pin F lowers T 1 1 1 1 1 r~ 5/1618 22 2 4 28 6/1 4 8 12 D A T E - l — 15 -f- 0 18 6 0 0 - Thrum Flowers 5/16 19 15 18 SEED SET / 94 FIGURE 24. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population BL4 during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. g 300 I" 200 £ 100 co o c O 1 1 1—I r 5/22 24 28 6/1 4 D A T E T — i — r — i — i — i — i — i — r 5/22 24 28 6/1 4 8 11 15 18 D A T E SEED SET / 95 FIGURE 25. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population PCI during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. i DATE O c o CL Thrum F l o w e r s 6/3 8 11 15 18 21 DATE SEED SET / 96 FIGURE 26. The number of seeds set per capsule in pins and thrums of Menyanthes trifoliata in population PC2 during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. u, 6 / 3 8 11 15 18 21 E D A T E ro i_ O co Q_ Thrum F lowers mpati i Cn R - 2 0 CL mpati i 4 0 0 -o u 3 0 0 - / X - 15-g o / V -> 2 0 0 - / ' - 10 n / 1 J 1 _ T3 1 0 0 - \ , - 5 v. E \ ' \ ; c /•N 0 R z u - 1 1 1 6 / 3 8 11 D A T E SEED SET / 97 FIGURE 27. The number of seeds set per capsule in pins and homostyles of Menyanthes trifoliata in population SLl during the flowering season of 1985. Solid lines represent average seed set per capsule calculated (1) with fruits that failed to set seed (heavy solid line) and (2) without failed fruits (lighter solid line). N= 10-20 capsules per floral morph for each collection date. Broken line represents average number of compatible pollen grains deposited on stigmas collected on the same dates that flowers were tagged for seed set data. N = 40 stigmas per floral morph for each collection date. Pin F l o w e r s CO I/) c flj 1_ O c o D. XJ 5/18 21 24 26 DATE Homosty le F l o w e r s - i — i — i i i 5/18 21 24 26 31 DATE SEED SET / 98 dates during the flowering season with the average number of compatible pollen grains received revealed that on some collection dates, some floral morphs received fewer than six compatible pollen grains per seed set. This indicates that seed set in Menyanthes may be limited by a lack of compatible pollinations during part of the flowering season. Pins from S L l were the only plants that received fewer than six compatible pollen grains per seed set during most of the flowering season. At S L l , pin and homostyle seed set was much higher in flowers that were hand-pollinated with compatible pollen than in naturally-pollinated flowers. Pin seed set in hand-pollinated flowers averaged 14.3 seeds per capsule (s.d. = 4.47) compared to an average of 5.8 seeds per capsule (s.d. = 2.12) in naturally-pollinated flowers. For homost3'les, seed set in hand-pollinated flowers averaged 11.7 seeds per capsule (s.d. = 3.26), whereas seed set in naturally-pollinated flowers averaged 5.1 seeds per capsule (s.d. = 2.79). Percentage seed set in pins was 47.5 percentage points higher with supplemental pollination averaging 81.6% (s.d. = 21.55) compared to an average of 34.1% (s.d. = 38.23) in naturally-pollinated flowers. For homostyles percentage seed set was 37.8 percentage points higher in hand-pollinated flowers averaging 69.6% (s.d. = 14.19) compared to an average 31.8% (s.d. = 31.23) in naturally-pollinated flowers. Percentage fruit set wras calculated from bulk samples of undehisced fruits. A t B L 2 , E L 3 , BL4 and S L l percentage fruit set was similar for both floral morphs in each population (Table 15). However, at P C I and PC2 percentage fruit set was 2-3 times greater in thrums than in pins. Percentage fruit set was TABLE 15. Seed se t . percentage f r u i t set and percentage seed set in p i n s , thrums and homostyles of Menyanthes trifol iat a from nine populations in southwestern B r i t i s h Columbia. Values are the mean and standard dev ia t ion ( in parentheses). Means are ca lcu la ted from bulk samples of undehlsced f r u i t s . F lora l Percentage Seed set Percentage Populat ion morph N f r u i t set per capsule seed set D i morph i c BL2 BL3 BL4 PC 1 PC2 SL1 p i n thrum p i n thrum p 1 n thrum p i n thrum pin thrum pin homostyle 248 258 306 241 173 332 726 943 709 120 436 472 83.4 78 . 3 40. 2 58.5 63.6 78.5 29. 1 64 . 1 17.9 55.8 46.8 38.6 9.6 (8.06) 10.5 (8.78) 6.1 (10.60) 8.2 ( 1 1 .28) 11.8 (12.27) 12.6 (9.78) 3.4 (7.29) 9.6 (10.78) 1.7 (5.54) 8.8 (10.79) 4.4 (6.71) 3.7 (5.61 ) 36.4 (29.51) 39.7 (30.33) 16.9 (27.71) 27.2 (31.00) 45.1 (41.11) 51.7 (37.02) 11.3 (23.59) 32.3 (33.64) 5.6 (17.13) 25.7 (29.78) 25.5 (36.41) 18.4 (27.67) Monomorph i c PC3 SM 1 SL2 thrum thrum homosty1e 1236 3225 232 18 .O 0. 1 2.0 0.79 (1.90) 0.01 (0.43) 0.05 (0.20) 3.9 (8.80) 0 . 2 ( 1 . 9 0 ) C O W w o CO w CO co S E E D SET / 100 highest at BL2 and lowest in the monomorphic population S M I . Within each population, seed set averages calculated from bulk samples of undehisced fruits (Table 15) were similar to seasonal seed set averages in tagged flowers (Table 14). Sample sizes of bulk seed set were much larger than those for seasonal seed set averages, and revealed some different relationships in the number of seeds set by the two floral morphs in each population. With bulk seed set averages, thrum seed set was significantly greater than pin seed set at BL3 (p<0.05) and P C I (p<0.001). On the other hand, there was no significant difference in seed set between pins and thrums from BL4. For the other three populations (BL2, PC2 and S L l ) , the relationships between the average number of seeds set by the two floral morphs were the same as those found in seasonal seed set averages. Percentage seed set calculated from bulk samples of undehisced fruits was also similar to seasonal percentage seed set values and averaged 27.9% (s.d. = 13.77) (Table 15). At BL3 , P C I and PC2, percentage seed set (calculated from bulk samples) was significantly higher in thrums than in pins (BL3: p<0.01, P C I : p<0.001, PC2: p<0.001). At S L l percentage seed set was significantly higher (p<0.05) in pins than in homostyles. A t BL2 and BL4 , there was no significant difference in percentage seed set between floral morphs. Bulk samples of undehisced fruits were also collected from three monomorphic populations: PC3, S M I and SL2 . Average seed set and percentage seed set in the monomorphic populations was much lower than that for the SEED SET / 101 same floral morphs growing in dimorphic populations (Table 15). A t PC3, SMI and SL2, seed set averaged less than one seed per capsule, and percentage seed set was less than 4%. D. DISCUSSION There are conflicting reports in the literature concerning pollen limited seed set in distjdous plants. In some species {Pulmonaria obscura (Ornduff 1979), Lithospermum caroliniense (Weller 1980), Cratoxylum formosum (Lewis 1982), Palicourea fendleri (Sobrevila et al 1983), Hedyotis caerulea (Ornduff 1980b) and Linum tenuifolium (Nicholls 1986)}, seed set is limited by a lack of compatible pollinations. But, in other species seed set appears to be limited by factors other than pollen availability. In populations of three distylous species {Amsinckia grandiflora (Ornduff 1976), Cryptantha flava (Casper 1983) and Primula elatior (Schou 1983)}, seed set was far below the potential maximum. However, the average number of compatible pollen grains deposited on stigmas of these species was higher than the number of ovules per ovary. Casper (1983) suggested that in Cryptantha flava, embryo abortion was responsible for the observed low seed set. In the six dimorphic populations of Menyanthes studied, seed set was always below the potential maximum. The percentage of seeds set per capsule (calculated from bulk samples of undehisced fruit) averaged 27.9% (s.d. = 13.77). Among populations, there was a very high correlation between the average number of seeds set by each floral morph during the whole flowering season and SEED SET / 102 the average number of compatible pollen grains deposited on stigmas. However, within most populations there was no correlation between seed set and pollen deposition on specific dates during the flowering season. Because both pollen deposition and seed set fluctuated so erratically during the season, the average number of compatible pollen grains received per seed set (on each collection date) was highly variable. Within each population, seed set appeared to be limited by pollen availability on some dates but not on others. But the lack of correlation between daily seed set patterns and pollen deposition patterns makes these results difficult to evaluate. Pins from SLl were the only plants in which there was a significant positive correlation between the daily seed set pattern and and the daily pollen deposition pattern. The average number of compatible pollen grains received per seed set in pins from SLl was less than six on four out of five dates measured. This suggests that seed set in these plants was limited by pollen availability. Results from hand-pollinations of pins and homostyles at SLl support this conclusion. Seed set in pin flowers from SLl that were hand-pollinated with large amounts of compatible pollen was 47.5 percentage points higher than seed set in naturally-pollinated flowers. Similarly, seed set in homostyle flowers from SLl that were hand-pollinated was 37.8 percentage points higher than seed set in flowers that were naturally-pollinated on the same days. Yet, comparison of daily seed set and pollen deposition patterns in homostyles revealed no apparent pollen limitation. In populations of Hedyotis caerulea (Ornduff 1980b) and Linum tenuifolium SEED SET / 103 (Nicholls 19S6), there was a considerable amount of variation in the number of compatible pollen grains deposited on individual stigmas. In Hedyotis caerulea, up to 21% of the stigmas held no pollen at all (Ornduff 1980b). In these two species, seed set was limited by a lack of compatible pollinations in some flowers, but in others there was ample pollen available for the fertilization of ovules. Although pollen loads of individual stigmas of Menyanthes were not examined, the erratic nature of daily pollen flow patterns during the season suggests that in Menyanthes there ma}' be a great deal of variation in the number of compatible pollen grains deposited on individual stigmas. Even though seed set data were collected from flowers pollinated on the same days as flowers collected for pollen flow data, the composition of their stigmatic pollen loads may have been very different. If seed set data were collected from flowers that received very few compatible pollen grains, and pollen flow data were collected from flowers that received an ample number of compatible pollen grains, pollen flow data would not correlate very well with seed set data. It is possible that in populations of Menyanthes, the lack of correlation between seed set and the number of compatible pollen grains received. on individual collection dates is partly a result of variation in the composition of individual stigmatic pollen loads. Results from hand-pollinations at SLl indicate that pollen availability may have had a greater influence on seed set in homostyles than daily patterns of pollen deposition and seed set suggest. Even though seed set in hand-pollinated flowers of Menyanthes was substantially higher than seed set in naturally pollinated flowers, maximum seed set was never obtained in supplementally pollinated flowers. The highest SEED SET / 104 percentage seed set that was obtained with hand pollination was 95%. However, percentage seed set in hand-pollinated pin flowers averaged 81.6% (s.d. = 21.55) and in homostyle flowers averaged 69.6% (s.d. = 14.19). Supplemental pollination has been found to increase seed set in two other distylous species, Lithospermum caroliniense (Weller 1980) and Palicourea fendleri (Sobrevila et al. 1983). In Lithospermum caroliniense, seed set was increased by 19 percentage points with supplemental pollination, but, as in Menyanthes, maximum seed set was never obtained. Because many factors have been shown to influence seed set in plants, it is probably unrealistic to think that only one factor limits seed set. In Menyanthes there is a strong correlation between the average number of compatible pollen grains received and average seed set, and in at least one population a lack of compatible pollen seems to be a major factor limiting seed set. However, it seems evident that factors other than pollen availability are influencing seed set in Menyanthes trifoliata. Comparison of seed set in monomorphic (SL2) and dimorphic (SLl ) populations of Menyanthes homostyles suggests that these homostyles are self-incompatible morphological homostyles, rather than self-compatible genetic homostyles. True genetic homostyles result from a crossover in the distyly supergene (Ganders 1979). Genetic homostyles are self-compatible because they have the stamens of one floral morph, and the pistils of the opposite floral morph. Long homostyles have thrum stamens and pin pistils, whereas short homostyles have pin stamens and thrum pistils. Homostyles from Stump Lake are morphologically equivalent to long homostyles. If they were recombinant homostyles, their stigmas would have the same compatiblity relationships as pin SEED SET / 105 stigmas. Pin pollen would be incompatible on their stigmas. Only pollen from thrum flowers or from their own anthers would be compatible on their stigmas, and self-pollination would result in substantial seed set. The proximity of homostyle stigmas to their own anthers would almost guarantee that some self-pollination would occur. Examination of stigmatic pollen loads of homostyles from S L l revealed extremely high numbers of own-form pollen grains. Presumably, homostyles from SL2, a population consisting entirely of homostyles, would receive similar numbers of own-form pollen grains. Yet, seed set in SL2 averaged 0.05 seeds per capsule (s.d. = 0.43). At S L l , where homostyles were growing with pins, homostyle seed set averaged 5.1 seeds per capsule (s.d. = 2.70). These results indicate that Stump Lake homostyles are self-incompatible and require intermorph pollen flow (pin pollen) for seed set to occur. Higher seed set in S L l homostyles following hand-pollination with large amounts of pin pollen supports this conclusion. Homostyle stigmas from Stump Lake have the same compatiblity relationships as thrum flowers. They are compatible with pin pollen and appear to be self-incompatible. These observations, combined with the observed morphological variation in Menyanthes flowers, support the view that the homostyles from S L l are not true genetic homostyles, but self-incompatible morphological homostyles that are equivalent to thrums with unusually long styles. Comparison of seed set in the pin/homostyle population at S L l to that in the similarly structured pin/thrum population at PC2 shows that thrums were more fecund than homostyles. Seed set in PC2 thrums averaged 16.0 seeds per SEED SET / 106 capsule (s.d. = 7.07), compared to an average of 5.1 seeds set per capsule (s.d. = 2.79) in S L l homostyles. These results agree with those found for thrums and homostyles of Mitchella repens (Ganders 1975a). In Menyanthes, there was a high correlation between the average number of compatible pollen grains deposited on stigmas and average seed set. Average seed set in homostyles was higher in flowers that were hand pollinated with large amounts of compatible pollen than in naturally pollinated flowers. This indicates that seed set in homostyle flowers was limited to some degree by a lack of compatible pollinations. The higher seed set in PC2 thrums corresponded to the receipt of a higher number of compatible pollen grains; PC2 thrums received 1.6 times more compatible pollen than homostyles. These results support Gander's (1979) hypothesis that morphological distyly is associated with greater fecundity, and demonstrate that higher seed set in thrums compared to homostyles resulted from thrums having received more compatible pollen grains than homostyles. VI. CONCLUSIONS This study was designed to investigate the distylous breeding system of Menyanthes trifoliata in order to determine (1) the effect of variation in morph frequency on pollen flow of a distylous plant, (2) whether morphological distyly increases the amount of compatible pollen deposited on stigmas of a diallelic self-incompatible plant, and if so (3) whether an increase in the amount of compatible pollen received is associated with an increase in fecundity. In two out of three populations examined (Beaver Lake and Pinecrest), flowers of Menyanthes trifoliata exhibited the typical floral dimorphisms associated with a distylous breeding system. Stigmas and anthers were reciprocally positioned in the two floral forms, and thrum pollen was significantly larger than pin pollen. Other floral characters such as anther length and petal length, dimorphic in some distylous species, displayed more interpopulational variation than intermorph variation in flowers of Menyanthes. A series of experimental pollinations revealed that in Menyanthes, morphological distyly is associated with a strong self-incompatibility system, and that an average of six compatible pollen grains are required to fertilize each ovule. The third population of Menyanthes studied (Stump Lake) consisted of pins and long homostyles. Although these homostyles were not tested for self-compatibility through self-pollination, other evidence suggests that they were self-incompatible morphological homostyles rather than self-compatible genetic homostyles. If the homostyles at Stump Lake resulted from a genetic crossover 107 CONCLUSIONS / 108 in the distyly supergene, their stigmas would have the same compatibility relationships as pin stigmas, and would be compatible with pollen from their own anthers (Ganders 1979). Examination of homostyle stigmatic pollen loads in Menyanthes revealed that they receive a high number of own-form pollen grains, probably as a result of self-pollination. If the homostyles were self-compatible, then the high number of own-form pollen grains deposited on stigmas should result in substantial seed set. Nevertheless, seed set in a population consisiting entirely of homostyles averaged 0.05 seeds per capsule (s.d. = 0.20). In comparison, homostyle seed set in the pin/homostyle population averaged 5.1 seeds per capsule (s.d. = 2.79); this indicates that intermorph pollen flow is required for seed set. Increased homostyle seed set following hand-pollination with pin pollen (11.7 seeds per capsule, s.d. = 3.26) supports the view that the homostyle stigmas have the same compatibility relationships as thrum stigmas, and are therefore self-incompatible morphological homostyles equivalent to thrums with unusually long styles. An average of 1% of the pollen produced in populations of Menyanthes trifoliata is deposited on conspecific stigmas. Comparison of seasonal pollen flow patterns in the six dimorphic populations of Menyanthes studied, revealed that in general, pollen flow is very erratic. Within each population, the size and composition of stigmatic pollen loads fluctuated considerabty during the flowering season. Pins and thrums experienced random, disassortative and assortative pollination at different times during the flowering season. The level of disassortative pollination varied from 16.7% - 65.9%. Homostyle stigmas were assortatively pollinated on all dates measured, probably as a result of the CONCLUSIONS / 109 proximity of their stigmas to their own anthers. Within each population, floral morph frequency (or more accurately, pollen morph frequency) had a very strong influence on the composition of stigmatic pollen loads averaged for the whole season. The percentage of compatible pollen deposited on stigmas of both floral morphs in each population (excluding homostyles) was highly correlated (r = 0.93, p< 0.001) with the percentage of compatible pollen produced in the population. However this relationship was not proportional. Within each population, the more abundant floral morph received slightly more compatible pollen than expected based on the freqeunc3' of pollen produced in the population. This is significant because it suggests that morphological distyly compensates for the rarity of one floral morph bj' increasing the proportion of compatible pollen received by the opposite floral morph. The actual number of compatible pollen grains deposited on stigmas did not correlate as strongly with pollen frequency in the population (r=0.69, p<0.05). This probably resulted in part from differences between floral morphs in the total number of pollen grains deposited on stigmas. Comparison of stigmatic pollen loads from the pin/homostyle population to those of thrums from a similarly structured pin/thrum population revealed that thrums received a much higher percentage of compatible pollen than the homostyles. Thrums received an average of 69.2% (s.d. = 31.60) compatible pollen, whereas homostyles received an average of 24.0% (s.d. = 30.30) compatible pollen. The difference between homostyles and thrums in the actual number of compatible pollen grains deposited on stigmas was not as great. Thrums received CONCLUSIONS / 110 an average of 1.6 times more compatible pollen than homostyles. However, the number of incompatible pollen grains deposited on homostyle stigmas was 4.4 times greater than the number received by thrums. Except for differences in style length, homostyle flowers were morphologically identical to thrum flowers. Both grew in populations with a similar morph frequencies. Therefore, the differences observed between thrums and homostyles in composition of stigmatic pollen loads can be attributed to the presence of the reciprocal stamen/style dimorphism in the pin/thrum population. In populations of diallelic self-incompatible plants where there are only two mating groups, the chances of a stigma receiving incompatible pollen are great. Darwin (1877) hypothesized that because of the reciprocal placement of stigmas and anthers in flowers of distylous plants, the amount of compatible pollen deposited on stigmas would be greater than the amount of compatible pollen received by a homomorphic diallelic self-incompatible plant. Results from my study of pollen flow in Menyanthes trifoliata provide some quantitative evidence in support of this hypothesis. In some populations of Menyanthes, morphological distyly increases the amount of compatible pollen deposited on stigmas, and reduces the amount of incompatible pollen received. Among the six dimorphic populations of Menyanthes studied, there was a very high correlation (r = 0.88, p< 0.001) between the average number of compatible pollen grains deposited on stigmas and the average number of seeds set. However, within each population there was little or no correlation between seed set and the number of compatible grains received on individual collection CONCLUSIONS / 111 dates during the flowering season. Pins from S L l were the only plants in which there was a significant positive correlation between the two (r = 0.83, p<0.05). In the six dimorphic populations of Menyanthes studied, seed set was always below the potential maximum. A n average of 27.9% (s.d. = 13.77) of the ovules in each ovary were fertilized. Comparison of the number of compatible pollen grains received to the number of seeds set revealed that on some dates during the flowering season, fewer than six compatible pollen grains were received per seed set. This indicates that, in populations of Menyanthes, seed set may be limited by pollen availability during part of the flowering season. However the lack of correlation between seed set and the number of compatible pollen grains received on individual dates during the flowering season makes evaluation of these results difficult. At S L l , pins and homostyles were hand-pollinated with large amounts of compatible pollen on the same dates in which seed set data were collected from naturally-pollinated flowers. Seed set in hand-pollinated flowers was an average of 42.7 percentage points higher than seed set in naturally-pollinated flowers indicating that seed set was limited by pollen availability in this population. For S L l pins these results are not surprising because comparison of their seasonal seed set and pollen deposition patterns revealed that they received fewer than six compatible pollen grains per seed set during most of the flowering season. For homostyles a similar comparison revealed that homostyle seed set was not limited by pollen availability. Results from hand-pollinations show that pollen availability may have had a greater influence on seed set in homostyles than the CONCLUSIONS / 112 comparison of seasonal pollen flow and seed set patterns indicate. I have suggested that in populations of Menyanthes trifoliata, the lack of correlation between seed set and the number of compatible pollen grains received on individual collection dates during the flowering season is partly a result of variation in the composition of individual stigmatic pollen loads. In addition, sampling techniques may have contributed to this lack of correlation. However, even with supplemental pollination, seed set in pins and homostyles from S L l never reached its potential maximum, and this suggests that in addition to pollen availabilty, other factors influenced seed set in populations of Menyanthes. Comparison of homostyle seed set in the pin/homostyle population to thrum seed set in the pin/thrum population showed that the higher number of compatible pollen grains, deposited on thrum stigmas was associated with higher fecundity. Thrums set an average of 10.9 more seeds per capsule than homostyles. Ganders (1974) hypothesized that morphological distyly increases the fecundity of a diallelic self-incompaitble plant by increasing the number of compatible pollen grains deposited on stigmas. Results from my stud}' of pollen flow and seed set in populations of Menyanthes provide quantitative evidence in support of this hypothesis. In populations of Menyanthes trifoliata, the reciprocal placement of stigmas and anthers in distylous flowers increases the amount of compatible pollen deposited on stigmas, and this increase in compatible pollen is associated with higher seed set. This investigation has provided considerable quantitative information on the influence of morphological distyly and variation in morph frequency on pollen flow CONCLUSIONS / 113 and seed set. In addition, this study is the first in which pollen deposition patterns have been examined at regular intervals during the flowering season. The methods used in this study have revealed the erratic nature of pollen flow in populations of distylous plants. In future studies, more emphasis should be placed on the investigation of this phenomenon in order to improve our understanding of the distylous mating system. BIBLIOGRAPHY Akamine, E.K., and G. Girolami. 1959. Pollination and fruit set in the yellow passion fruit. Tech. Bull. Hawaii Exp. Stn. No. 39. Baker, H.G. 1959. The contribution of autecological and genecological studies to our knowledge of the past migrations of plants. Amer. Nat. 93: 255-272. Barrett, S .C.H. 1977. Tristyly in Eichhornia crassipes (Mart.) Solms (Water Hyacinth). Biotropica 9: 230-238. 1978. Heterostyly in a tropical weed: the reproductive biology of the Turnera ulmifolia complex (Turneraceae). Can. J . Bot. 56: 1713-1725. 1980. Dimorphic incompatibility and gender in Nymphoides indica (Menyanthaceae). Can. J . Bot. 58: 1938-1942. Bertin, R.I. 1982. Floral biology, hummingbird pollination and fruit production of trumpet creeper (Campsis radicans, (Bignoniaceae). Amer. J . Bot. 69: 122-134. Casper, B. 1983. The efficiency of pollen transfer and rates of embryo initiation in Cryptantha flava (Boraginaceae). Oecologia 59: 262-268. 1985. Self-compatibility in distylous Cryptantha flava (Boraginaceae). New Phytol. 99: 149-154. Darwin, C. 1862. On the two forms, or dimorphic condition, in the species of Primula, and on their remarkable sexual relations. J . Proc. Linn. Soc. (Bot.) 6: 77-96. 1877. "The Different Forms of Flowers on Plants of the Same Species". John Murray. London. Dowrick, V . P . J . 1956. Heterostyly and homostyly in Primida obconica. Heredity 10: 219-236. Erdtman, G. 1960. The acetolysis method. Svensk. Bot. Tidskr. 54: 4. 114 / 115 Fernald, M.L. 1950. "Gray's Manual of Botany". 8th ed. American Book Co. New York. Ganders, F.R. 1974. Disassortative pollination in the distylous plant Jepsonia heterandra. Can. J . Bot. 52: 2401-2406. 1975a. Mating patterns in self-compatible distylous populations of Amsinckia (Boraginaceae). Can. J . Bot. 53: 773-779. 1975b. Fecundity in distylous and self-incompatible homostylous plants of Mitchella repens (Rubiaceae). Evol. 29: 186-188. 1976. Pollen flow in distylous populations of Amsinckia (Boraginaceae). Can. J . Bot. 54: 2530-2535. 1979. The biology of heterostyly. New Zealand J . Bot. 17: 607-635. Glover, D.E., and S.C.H. Barrett. 1986. Stigmatic pollen loads in populations of Pontederia. cordata from the southeastern U.S. Amer. J . Bot. 73: 1607-1612. Griffiths, A . J . F . , and F.R. Ganders. 1983. "Wildflower Genetics". Flight Press. Vancouver. Guth, C .J . , and S.G. Weller. 1986. Pollination, fertilization and ovule abortion in Oxalis magnifica. Amer. J . Bot. 73: 246-253. Hewett, D.G. 1964. Menyanthes trifoliatia L. J . Ecol. 52: 723-735. Hitchcock, C.L. , and A . Cronquist. 1973. "Flora of the Pacific Northwest". Univ. of Washington Press. Seattle. Knuth, P. 1899. "Handbuch der Blutenbiologie". 11,2: Verlag urn Wilhelm. Engelman, Liepzig. Levin, D.A. 1968. The breeding system of Lithospermum- caroliniense: Adaptation and counteradaptation. Amer. Nat. 102: 427-441. Lewis, D. 1982. Incompatibility, stamen movement and pollen economy in a / 116 heterostyled tropical forest tree, Cratoxylum formosum (Guttiferae). Proc. R. Soc. Lond. 214: 273-283. Lloyd, D.G. 1965. Evolution of self-compatibility and racial differentiation in Leavenworthia (Cruciferae). Cont. Gray. Herb. 195: 3-134. McDade, L.A. 1983. Pollination intensity and seed set in Tricanthera gigantea (Acanthaceae). Biotropica 15: 122-124. Neisess, K.R. 1984. Heterostyly in Salvia brandegei (Lamiaceae). Madrono 31: 252-254. Nicholls, M.S. 1985. Pollen flow, population composition and the adaptive significance of distyly in Linum tenuifolium (Linaceae). Biol. J . L inn. Soc. 25: 235-242. 1986. Population composition, gender specialization, and the adaptive significance of distyly in Linum perenne (Linaceae). New Phytol. 102: 209-217. Nilsson, S. 1973. Menyanthaceae Dum. Taxonomy by Robert Ornduff. World pollen and spore flora. 2: 1-20. Oleson, J . 1979. Floral morphology and pollen flow in the heterostylous species Pulmonaria obscura (Boraginaceae). New Phytol. 82: 757-767. Ornduff, R. 1966. The origin of dioecism from heterostyly in Nymphoides (Menyanthaceae). Evol. 20: 309-314. 1970a. Heteromorphic incompatibility in Jepsonia malvifolia. Bul l . Torrey. Bot. Club 97: 258-261. 1970b. Incompatibility and pollen economy in Jepsonia parryi. Amer. J . Bot. 57: 1036-1041. 1971. The reproductive system of Jepsonia heterandra. Evol. 25: 300-311. 1975. Heterostyly and pollen flow in Hypericum aegypticum (Guttiferae). / 117 Bot. J . Linn. Soc. 71: 51-57. 1976. The reproductive system of Amsinckia grandiflora- a distylous species. Syst. Bot. 1: 57-66. 1978. Features of pollen flow in dimorphic species of Lythrum section Euhyssopifolia. Amer. J . Bot. 65: 1077-1083. 1979. Pollen flow in a population of Primula vulgaris. Bot. J . Linn. Soc. 78: 1-10. 1980a. Pollen flow in Primula veris. PI. Syst. Evol. 135: 89-93. 1980b. Heterostyly, population composition and pollen flow in Hedyotis caerulea. Amer. J . Bot. 67: 95-103. 1982. Heterostyly and incompatibility in Villarsia capitata (Menyanthaceae). Taxon 31: 495-497. 1986. Comparative fecundity and population composition of heterostylous and non-heterostylous Villarsia (Menyanthaceae) in Western Australia. Amer. J . Bot. 73: 282-286. Philipp, M., and O. Schou. 1981. Distyly, self-incompatibility, pollen loads and fecundity in Anchusa officinalis (Boraginaceae). New Phytol. 89: 693-703. Pielou, E.C. 1961. Segregation and symmetry in two-species populations as studied by nearest-neighbour relationships. J . Ecol. 49: 255-269. Price, S.D., and S.C.H. Barrett. 1984. The function and adaptive significance of tristyly in Fontederia cordata L. (Pontederiaceae). Biol. J . Linn. Soc. 21: 315-329. Rama Swamy, and Bir Bahadur. 1984. Pollen flow in dimorphic Turnera subulata. New Phytol. 98: 205-209. Reddy, N., and' Bir Bahadur. 1976. Heterostyly in Nymphiodes indica (L.) Griseb. J . Indian Bot. Soc. 55: 133-140. / 118 Richards, A . J . , and H. Ibrahim. 1978. Estimation of neighbourhood size in two populations of Primula veris. In "The Pollination of Flowers by Insects". Edited by A . J . Richards. London. Academic Press. 165-174. Schemske, D.W., and C. Fenster. 1983. Pollen grain interactions in a neotropical Costus: effects of clump size and competitors. In "Pollen: Biology and Implications for Plant Breeding". Edited by D.L. Mulcahy and E. Ottaviano. Elsevier Scientific Publishing Co. Amsterdam, pp. 405-510. Schou, O. 1983. Distyly in Primula elatior (Primulaceae), with a study of flowering phenology and pollen flow. Bot. J . Linn. Soc. 86: 261-274. Shore, J . , and S.C.H. Barrett. 1984. The effect of pollination intensity and incompatible pollen on seed set in Turnera ulmifolia (Turneraceae). Can. J . Bot. 62: 1298-1303. Silander, J .A . , and R.B. Primack. 1978. Pollination intensity and seed set in the evening primrose Oenothera fruticosa. Amer. Midi. Nat. 100: 213-216. Snow, A .A . 1982. Pollination intensity and potential seed set in Passiflora vitifolia. Oecologia 55: 231-237. Sobrevila, C , Ramirez, N . and N. Enreeh. 1983. Reproductive biology of Palicourea fendleri and Palicourea petiolaris (Rubiaceae), heterostylous shrub of a tropical cloud forest in Venezuela. Biotropica 15: 161-169. Sokal, R.R., and F . J . Rohlf. 1981. "Biometry". 2nd edition. Freeman. San Francisco. van Dijk, W. 1943. L a decouverte de l'heterostylie chez Primula pas Ch. de l'Ecluse et P. Reneaulme. Nederlandsch Kruidkundig Archief 53: 81-85. Warming, E. 1886. Om b3^gningen og den formodede best ovningsmaade af nogle gronlandske blomster. Oversigt ober Det. Kgl. danske Vid. Selsk. Forh. pp. 1-59. Weller, S. 1980. Pollen flow and fecundity in populations of Lithospermum caroliniense. Amer. J. Bot." 67: 1334-1341.