Oecologia (1992) 89:277-283 Oecologia �9 Springer-Verlag 1992

Gene flow inferred from seed dispersal and pollinator behaviour compared to DNA analysis of restriction site variation in a patchy population of Lotus corniculatus L.

I.R. Rasmussen 1 and B. Brodsgaard 2

1 Institute of Plant Ecology, University of Copenhagen, Oster Farimagsgade 2D, DK-1353 Copenhagen K, Denmark 2 Institute for Biological Chemistry B, University of Copenhagen, Solvgade 83, DK-1307 Copenhagen K, Denmark

Received April 26, 1991 / Accepted in revised form July 14, 1991

Summary. Gene flow was investigated in a natural pop- ulation of Lotus corniculatus L. (Fabaceae) using a combination of pollen and seed dispersal studies and a recombinant DNA technique. The population is spatially heterogeneous and grows with Empetrum nigrum. L. cor~ niculatus is pollinated by the pollen-collecting bumblebee Bombus lapidarius L. Most pollinator flights occurred within patches, as bees usually visit nearest-neighbour plants, show no marked directionality, and forage mostly within patches. Gene flow by seeds is also limited, rein- forcing the pattern of gene flow within patches. However, 2.6% of pollinator flights are between patches and con- siderable pollen carryover also occurs. Thus, gene flow between patches is potentially sufficient to retard or prevent genetic differentiation in spite of the patchy sub- structuring of the population. A sub-set of the popula- tion was analysed for restriction fragment length poly- morphisms (RFLPs) to document the actual gene flow pattern of the population. The DNA analysis revealed significant levels of genetic differentiation between the patches. The level of gene flow that can be inferred from the distribution of genetic variation is surprisingly re- stricted, as compared to gene flow inferred from pollina- tor behaviour, and emphasizes that stochastic processes like genetic drift and founder effects may have a strong impact on the prevailing genetic structure.

Key words" Lotus corniculatus L. - Patchiness - Gene flow - Pollinator behaviour - Genetic variation

Gene flow in natural plant populations is determined by seed and pollen flow. Except in species where the seed dispersal mechanism is more specialized, e.g. mediated by water (Waser et al. 1982) or animals (Beattie 1978), the seed flow component of gene flow is of minor impor- tance relative to the pollen flow component (Levin and

Offprint requests to." I.R. Rasmussen

Kerster 1968; Campbell and Waser 1989). Consequently, seed dispersal is frequently left out in investigations of gene flow (but see Levin and Kerster 1968; Levin and Kerster 1969a; Schaal 1980).

In the pollen flow component of gene flow in insect- pollinated plants, the pollinators and their movement through the plant population have a profound influence on the breeding structure as well as on the genetic struc- ture of the population (Levin 1978, 1983; Schmitt 1983). The pollinators are, however, in their turn, influenced by the spatial structure of the plant population because plant density and distribution affect the movements of pollinators and thereby the dispersal pattern of pollen (Levin and Kerster 1969a, b; Schaal 1978; Schmitt 1983).

Here we examine seed and pollen flow in a spatially patchy population of the insect-pollinated Lotus cor- niculatus L. in which the pollinators are pollen-collecting bumblebees. Our first aim is to examine seed dispersal and to infer pollen flow in the population by observation of pollinator behaviour and by experimental manipula- tion in a pollen carryover experiment. Secondly we exam- ine the genetical variation in the population and relate it to the results of the study of seed and pollen flow. In previous studies the pollen component of gene flow has most often been inferred from pollinator flight distances alone. The conclusions from such data may, however, diverge from conclusions based on the use of genetic markers. Genetic markers, either isozyme markers (Schaal 1980; Handel 1982; Smyth and Hamrick 1987; Ellstrand et al. 1989) or DNA markers (Learn and Schaal 1987; Schaal et al. 1987; Nybom and Schaal 1990) in most cases provide a more accurate measure of gene flow.

We have chosen a system of patchy individual distri- bution and pollen-collecting pollinators, primarily be- cause investigations in spatially heterogeneous habitats are few (Hodges and Miller 1981; Sih and Baltus 1987) and most studies in such habitats have examined systems in which the insects forage for nectar and not for pollen (Heinrich 1979; Plowright and Galen 1985; Sowig 1989). Furthermore, the patchiness of the population facilitates observations of pollinator behaviour, thereby creating a

278

u n i q u e o p p o r t u n i t y o t c o m p a r e resul t s o f the t r a d i t i o n a l o b s e r v a t i o n a l t e c h n i q u e s w i t h t he m o r e r ecen t l y de- v e l o p e d r e c o m b i n a n t D N A t echn iques .

Material and methods

Plant material and study area

L. corniculatus is a weedy herbaceous perennial legume with one to several prostrate or erect shoots. Flowering is indeterminate and in Denmark takes place between late May and late September. The inflorescences are umbels consisting of 2-5 yellow bisexual flowers. The species is self-incompatible (Jones and Turkington 1986) and is, at our study site, pollinated by the large bumblebee Bombus lapidarius L. foraging for pollen. Other Bombus species and species of syrphid flies were occasionally observed near or on L. cornicula- tus, but B. lapidarius was the only insect species foraging on the flowers and served as the only effective pollinator.

The study was carried out in the dune heathlands of Korshage near the coast of Kattegat, North Zealand, Denmark and at the experimental field of the Botanical Garden of Copenhagen at Tgts- trup, 30 km west of Copenhagen. The Korshage area experiences a continued uplift and the subsequent colonization and plant suc- cession has been thoroughly investigated by B6cher (t952) and it is apparent, from the present species composition and the patchy species distribution in the area, that the succession of the area started approximately 100 years ago. The area has now reached a stable stage of the succession with a mosaic of patches of Empetrum nigrum L., most of which are associated with L. corniculatus, inter- spersed with dry areas dominated by mosses and lichens and the low grass, Corynephorus canescens (L.) Beauv., without L. eornieulatus. The L. eorniculatus population has been present for approximately 20 years but in recent years a notable increase in the number of individuals has been observed.

At Korshage six patches were selected that were characteristic for the study site with respect to both size and interpatch distance. Of these, three patches (A, B and C) were investigated for pollen flow, inferred from pollinator movements, whereas the genetic variation was investigated in all six patches.

The field observations were conducted from 20 June to 10 July, 1988 and from 19 June to 9 July 1989. The weather during these periods was similar in the two years, being sunny, dry and with a temperature of 25-30 ~ C. Data were collected between 5 a.m. and 8 p.m. with special emphasis on the period between 9 a.m. and 5 p.m. when the bees were most active.

Seed dispersal experiments

The seeds of L. corniculatus are dispersed by ejection on the ex- plosive dehiscence of the pod (Jones and Turkington 1986). Seed flow was investigated by allowing four plants to disperse their seeds onto a bare field at the experimental garden of T~strup in the early autumn of 1988. Each of the plants were allowed a free circum- ference of 7 m. The number of seedlings and their distance from the source plant were recorded the following spring.

Pollinator flight distance and directionality

At Korshage a rectangular area of 75 x 55 m was chosen within which each plant was mapped and located by coordinates, using land surveying equipment. Each day the flowering individuals were marked with numbered flags, 20 cm above ground, and flower-visit- ing bees were followed through the population and their flight distances, sequence of visits and directionality were recorded. We used a portable tape recorder to record the sequence of visits in a foraging bout. A foraging bout started when a bee began foraging

in the observed patch and ended when the bee shifted to the only other abundant flowering species in the area, Jasione montana L., or left the area. Flight distances, to the nearest 5 cm, were measured in the field as the linear distance between two successively visited plants. The directionality, i.e. the turning angles between arrival and departure directions, was marked on the maps with coordinates, and the angles were measured for all foraging bouts where 5 or more plants were visited.

Pollen carryover experiments

The pollen carryover experiments took place in the experimental garden in a field of planted L. corniculatus using fluorescent dye as a pollen analogue. Each plant was surrounded by six neighbouring plants giving a hexagonal array with 1.5 m spacing of plants. Before each run of the experiment, 30-40 recently opened flowers on a plant in the middle of the array were loaded with dye grains (dye no. R-18, 'magenta' supplied by RADIANT COLOR, Houthalen, The Netherlands). After an initial visit on the dye-loaded plant the bees were followed during their foraging bouts through the experi- mental population. The amount of dye present on the source flowers was examined between runs and flowers were refilled when they were depleted. During each run the visited inttorescences were marked and the number of inflorescences visited on each plant was recorded. When a run was terminated, the bumblebee was caught and kept confined until the experiments of that day were over, to prevent uncontrolled contamination with dye. Between runs the dye-loaded plant was kept covered to prevent uncontrolled spread of dye.

After a run, the visited inflorescences on each plant were picked and stored in separate, numbered plastic boxes at 5 ~ C. Later the same day each flower in the inflorescences was examined for occur- rence of dye on the stigma, using a 50 x dissecting microscope. The presence or absence of dye on the plants in the sequence of visits was then established. We determined only the presence of dye, since we found that counts of number of dye grains were not precise due to the varying size of the dye grains and their tendency to form small lumps.

Analysis o f genetic variation with the restriction fragment length polymorphism (RFL P) method

In 1988 there was a total of 139 plants distributed in the six patches. DNA was extracted from 30 of these and the genetic diversity was investigated with a recombinant DNA technique analysis of restric- tion fragment length polymorphism (RFLP) (Brodsgaard and Ras- mussen 1990). The RFLP technique takes advantage of the fact that DNA can be digested with restriction enzymes that recognize specif- ic D N A sequences and cut the DNA at these restriction sites. The resulting fragments of DNA can be induced to migrate according to size in an electric field on an agarose gel. After transfer of the DNA onto nylon membranes, a radioactive probe is hybridized with the specific homologous fragments. Subsequently these frag- ments are distinguished on a photographic film, an autoradiograph. Polymorphisms, i.e. differences in DNA fragment lengths, arise when the restriction enzymes cut the DNA at different sites due to changes in the DNA sequence. We have followed the principles described in many of the recent papers on DNA analysis of plants (e.g. Helentjaris et al. 1985; Schaal et al. 1987); however, we have used species-specific probes from a random gene-library construc- ted by cloning total cellular L. corniculatus DNA into the vector pTZ18U, using the restriction enzyme EcoRI (Mead et al. 1986; Sambrook et al. 1989). Two probes, the sizes of which are 2.1 kil- obases and 3.3 kilobases respectively, were hybridized to DNA digested with four different restriction enzymes (EcoRI, EcoRV, HindIII and HaeIII). The probes were tested by hybridization experiments and found to be of nuclear origin. DNA extractions were done by the procedures of Rogers and Bendich (1988) and

Jofuku and Goldberg (1988). The methods for plasmid isolation, restriction endonuclease digestion, ligation of DNA fragments, and transformation of plasmid DNA, as well as the procedures asso- ciated with the southern blot procedure are described by Sambrook et al. (1989).

Results

Seed dispersal

A total of 1172 seedlings originating from four plants were scored, with 240 from plant 1, 110 from plant 2, 239 from plant 3 and 583 from plant 4. The distances of seedlings dispersed from plants 1, 2 and 3 were identical, when tested for equal distributions by a Kolmogorov- Smirnov two-sample test (Sokal and Rohlf 1981) which allowed pooling of the data from these plants. The fre- quency distribution curve (Fig. 1) was highly leptokurtic,

~6

30

20

10

150 3 0 0 4 5 0

Dispersal distance of seedl ings (cm)

Fig. 1. Pooled data on the distribution of seedlings dispersed from plants 1, 2 and 3, measured as distances from seedlings to their maternal source (77 = 589)

279

with 80% of all seedlings within a radius of 2 m from the mother plant, but dispersal distances up to 4.95 m were recorded. The distances of seedlings dispersed from plant 4 deviated slightly from the other three plants, when tested for equal distributions by a Kolmogorov-Smirnov two-sample test. The frequency distribution curve is also highly leptokurtic but more skewed to the right and the longest dispersal distance was 5.70 m. A greenhouse ex- periment (Brodsgaard and Rasmussen 1990), measuring the potential dispersal distances of seeds, showed com- parable maximum dispersal distance and a distribution of dispersal distances similar to the one observed for seedlings in the field. We can therefore assume with great confidence that the distribution of seedlings accurately reflects the dispersal distances of the seeds. The green- house experiment furthermore showed that on average approximately 5% of all dispersed seeds were found as seedlings.

Pollinator flight distances

In 1988 and 1989, a total of 465 flights made by 74 bees were measured. The number of plants visited in each foraging bout as well as the time spent per patch varied among bees. Of all flights recorded, only 2.6% were between patches, whereas the rest occurred within the patches. The frequency distribution of flight distances within patches (Fig. 2) was leptokurtic with a long 'tail' and corresponded on the whole with what other authors have found (Levin and Kerster 1969a, b; Pyke 1978; Heinrich 1979; Schaal 1980; Schmitt 1980; Waddington 1981; Olesen and Warncke 1989). The maximum dis- tance recorded was 7.85 m and the most common range was 0.46-0.60 m, which coincides well with the most common distance between a flowering plant and its nearest flowering neighbour (Fig. 2, insert). This shows that B. lapidarius most often visited the nearest neigh- bour plant (Fig. 3) followed by flights to second and third nearest neighbour. Flights to plants further away than

s "6

15

10

60

"5

120 180 240 300 360

40J

3O

20 i

60 120 180 240

Distance (cm) to nearest neighbour

m , - ~ m . . , q m q q m 420 480 540 600 660

Pol l inator fl ight distance (cm)

Fig. 2. Distribution of flight dis- tances of Bombus lapidarius on plants of Lotus eornieulatus within patches (n = 465). Insert: Distribu- tion of nearest-neighbour dis- tances in the population (n= 139)

280

"6 o~

25 -

20 -

15 -

10 -

5 -

0 I I I I , 2 3 6 7 8 9 >

I 4 5

N e a r e s t n e i g h b 0 u r

Fig. 3. Percentage distribution of pollinator flights to nearest (1), ninth nearest (9) and to more distant neighbours (> 9), (n = 417)

20

15 iiiiiiiiii!i[i!ili!iiiil

Eiiii!iiiiiii!~iiiiiiill !iiiiiiiiiiii[i!iiiiiiii

,~ 10 ~iiiiiiiiiiiiiiiiii!iiiii iiiiiii!!iiiiiiiiiiiiiiiii

5 iiiiiiiiiii!iiiiiiiiiiii! iiiiiii~!ii!!iiiiiiiiiiiii

~iii ii iiiiiii iii!iiiiiiiiiiiiiiiiiiiii

iiiii!;iiiz~ililili:;iil

0 .

_ "7 "7 ~ ~'3 N "7 "7

+ I

Turning angle (degr.)

Fig. 4. Distribution of angle between departure and arrival direc- tions (in degrees) for Bombus lapidarius foraging for pollen on Lotus corniculatus (n = 342)

the ninth nearest neighbour, which often took the polli- nator across an entire patch, were the fourth most com- mon behaviour.

Pollinator flight directionality

turns (only in 168 of the 342 cases expected at perfect alternation), which would have created an overall pattern of directionality.

Pollen carryover

In a total of 16 experimental runs 371 inflorescences on 88 plants were visited. Of these, 164 inflorescences on 61 plants had dye 'deposits'. The frequency of plants with dye deposited on stigmas declined with plant sequence number (Fig. 5). The bumblebees visited unequal num- bers of inflorescences on the different plants, which caused a higher chance of dye occurrence on some plants; however, this bias occurred randomly with re- spect to sequence number, and should not affect the results.

Dye was transported up to the l lth plant in the sequence (not shown in Fig. 5 because data on dye trans- port to plants 9, 10 and 11 were based on one observation only). This corresponded to a transport across 40 in- florescences or 64 flowers. Thus, the pollen carryover experiments reveal a remarkably long dispersal. Further- more, in several runs dye was found even on the last flowers visited by the bee before capture, indicating that the tail of the deposition curve has not been fully assess- ed.

Genetic variation with the RFLP method

The RFLP analysis (Fig. 6) uncovered a high level of genetic variation in the population at Korshage (Brodsgaard and Rasmussen 1990). Three separate poly- morphisms were revealed. One probe detected a varia- tion in EcoRI restriction sites, with some individuals lacking a restriction site present in others (Fig. 7). The other probe also detected a restriction site variation, using the enzyme EcoRV, and furthermore revealed a deletion/insertion event seen with all enzymes used. It is here treated as a deletion, i.e. a piece of the DNA strand

The distribution of departure relative to arrival angles was significantly different (P < 0.01) from a uniform dis- tribution (Fig. 4) (Z 2= 21.76, n= 342). At the edge of a patch the bees had to choose between leaving the patch or turning back into it. If the turning angles measured from the edge plants, defined as plants with half of their circumference free of other flowering plants, are excluded from the computations, the difference from a uniform distribution is further strengthened, yielding a probabil- ity of P < 0.001 (;~2 __ 33.37, n = 291). This shows that the edge plants affect the flight pattern of the bees, provoking more backward flights than are observed in the middle of the patches. When reaching the edge of a patch the bees turned back into the patch in 67 % of the cases and continued foraging; alternatively they shifted to another species altogether or they took long flights to other parts of the population. The bees did not turn to the left more often than to the right (P> 0.1) (•2= 1.42, n= 342), and they did not show any alternation between right and left

0 E 0 Q- 0

0 t -

8, O

O

1,0

0,8

0,6

0,4

0,2

0,0 �9

o o

i

o

i �9 i , i �9 i �9 i , i �9 i , i , i

0 1 2 3 4 5 6 7 8

P lan t s e q u e n c e n u m b e r

Fig. 5. The frequency of plants with dye grains present for each plant sequence number. The number of observations (n) declines gradually from 16 for the first plant sequence number to 3 for the last sequence number

* / 1

" // i I

, I / ~ D

* , t , 1

"

/ / / . / / !

i" " i i

// �9 / / �9 / '

�9 / �9 a

�9 / t i

/ , �9 /

i i

�9 I I �9 / I �9 I / �9 n n

* l l * I I * / I �9 i i

A

O 1 A 3 2

1 212 1l �9 olAl*J

/ /

t I I I I I I I [ I

I 15m

Fig. 6. Map of the population of Lotus corniculatus, with patches A to F, showing the summary of RFLP patterns (compound geno- types) of the 30 individuals investigated. Map: Small dots = plants not investigated, full lines = patch edges, broken lines = paths through the area, asterisks = fence. Genotypes: Solid rectangle: - EcoRI and - Eco RV restriction sites, -- deletion; Solid triangle: - EcoRI and + EcoRV restriction sites, -- deletion; Open triangle: - EcoRI and + EcoRV restriction sites, + deletion; Solid circle: +EcoRI and -EcoRV restriction sites, -deletion; Open circle: +EcoRI and -EcoRV restriction sites, +deletion. Insert." The number of mutations or genomic changes necessary to advance from one compound genotype to another

281

missing entirely in some individuals. Within the patches the individuals were either genetically homogeneous or separated by one muta t ion event only (Fig. 6, insert). Between closely situated patches the individuals were generally separated by one or two mutat ion events. Be- tween the more distant patches the separation was more pronounced with up to three muta t ion events. Patch C seemed to occupy a transitional position, being closely related to both patches D and E and the group of patches A, B and F. The probabil i ty of finding the observed distribution of genotypes by chance alone is extremely small (P < 0.0004) (Fisher exact test on D N A data; Sokal and Rohl f 1981), which strongly suggests that genetic differentiation between patches is considerable.

Discussion

The distribution of seed dispersal distances for L. cor- niculatus, with the majori ty of the dispersal events within 2 m f rom the source plant, resembles dispersal patterns reported in other studies of species with ballistic seed dispersal (e.g. Levin and Kerster 1968; Schaal 1980; Stamp and Lucas 1983). Gene flow mediated by seeds at Korshage will probably be even more restricted than in the experimental garden because the dense cover of E. nigrurn impedes the free dispersal of the seeds. Further- more, the correlation between the distribution of L. cor- niculatus and E. nigrum implies that, in order to con- tribute to the actual gene flow, any long-distance dis- persed seeds must reach a patch of E. nigrum and estab- lish there. We do not know how frequent such long-dis- tance transportat ions of seeds are, but we believe that such events are very rare.

The pollinators, on the other hand, have no difficulty in crossing the 10-40 m f rom patch to patch. In spite of this, most flights are within the patches. They move f rom one patch to another in only 2.6% of their flights. Their staying in the patches conforms with theories of energet- ically optimal behaviour (Heinrich 1979; Z immerman 1982), since the pollen of L. corniculatus is not substan-

Fig. 7. Autoradiograph showing an example of the polymorphismsdetected by the RFLP analysis. The DNA was digested with the restriction enzyme EcoRI, run by electro- phoresis in an agarose gel, transferred to a nylon membrane and hybridized to a spe- cies-specific probe. Due to changes in the DNA sequence the individuals in lanes 1-10 lack the restriction site that, in the remain- ing individuals, divides the DNA fragment in two. The weaker intensity of the smallest bands can be attributed to the probe only binding to a small part of the fragment. Lanes 1-3: individuals from patch A; lanes 4-7+ 9-10: individuals from patch B; lanes 11-16: individuals from patch E; lanes 8 + 17." size marker (lambda DNA digested with BstEII). The size of the fragments is in- dicated on the picture; the size of the probe is 3.3 kilobases (kb)

282

tially depleted after a visit of a bee. Thus, in contrast to plants which are foraged for nectar (Levin et al. 1971; Pyke 1978; Woodell 1978; Dreisig 1985), L. corniculatus flowers can be visited several times and yet give a high reward in terms of pollen to the bee. Furthermore, the lack of overall directionality of flights, and the fact that the bees more often returned back into the patch than continued to another one when they approached the patch edges should minimize energy expenditure. At the same time this behaviour reduces the exchange of pollen between patches.

The pollen dispersal distance is determined not only by the flight behaviour of the pollinators but also by the magnitude of pollen carryover (Levin 1981), which varies greatly between species (Thomson and Plowright 1980; Waser and Price 1983 ; Galen and Plowright 1985; Svens- son 1985; Thomson and Thomson 1989). For the system of L. corniculatus and B. lapidarius, dye from the source plant was recovered on plants with high sequence num- bers. Caution is necessary, however, when applying results from an experiment with dye as a pollen analogue to natural conditions. Nonetheless, inferred from the carryover of dye it seems probable that pollen carryover can contribute substantially to the gene flow between patches.

From the data on seed dispersal and pollinator behav- iour we conclude that the short distance of seed dispersal, and the tendency of the bees to remain in the patches, will restrict gene flow between patches of L. corniculatus. The substantial carryover of dye, on the other hand, apparently counteracts this pattern. It has been suggested that even very low levels of migration (as low as one individual per generation) between populations are suf- ficient to prevent significant genetic differentiation in the absence of selection (Levin 1984; Slatkin 1987). In our system, the 2.6% of flights from patch to patch and the high carryover suggest a gene flow exceeding this mini- mum level for isolation considerably, and we would not expect the population to be genetically substructured.

The level of gene flow that can be inferred from the DNA analysis of the genetic variation is, however, sur- prisingly restricted. Previous studies on pollen flow in other systems using isozymes as genetic markers have, however, shown an actual gene flow that is greater than that inferred from pollinator movements alone (Schaal 1980; Handel 1982). This discrepancy can be interpreted in several ways. To begin with, natural selection could have created, or at least helped to maintain, an initial genetic substructuring of the population (Nevo et al. 1986; Slatkin 1987). However, all patches in the study area appear to experience similar conditions, and though no investigation has been carried out, we consider it unlikely that selection alone explains the discrepancy between the two measures of gene flow.

Another explanation is that stochastic processes, such as genetic drift and founder effects, may be important within the population. Genetic drift has been suggested as a predominant force in structuring a population of Clematis fremontii (Learn and Schaal 1987) showing a genetic subdivision comparable to that of L. eornieulatus in our study.

Founders of a population and their initial genetic

variation are well known to strongly affect the amount of subsequent variation (Learn and Schaal 1987; Levin 1988). From the previous studies in the area (B6cher 1952) it is apparent that the L. corniculatus population has been established relatively recently, probably within the last 20 years. Each of the patches at Korshage may then have been initiated by a few founders, and the present distribution of genetic variation could, conse- quently, be a reflection of past recruitments. Also, a long lived seed bank (Milberg 1990) within the patches could act as a stabilizing factor counterbalancing influx of new genotypes. Furthermore, as the RFLP analysis was done on mature individuals in 1988, there is a time scale dif- ference between the gene flow as we infer it from the seed dispersal and pollinator behaviour in 1988 and 1989 and the gene flow that we deduce from the DNA analysis. We believe, therefore, that the the discrepancy between the genetic structure of the population and the gene flow patterns inferred from pollinator behaviour can most probably be explained by stochastic effects of recent colonization, but that these effects will be disappearing in the near future due to exchange of pollen between the subpopulations. This conclusion on gene flow at Kor- shage could not have been reached if pollinator behav- iour or restriction site variation had been analyzed separately, and shows that a combination of the two approaches can yield information about populations that cannot be obtained otherwise.

Acknowledgements. We are grateful to Sven Jonasson, Hans Erik Svart Madsen, Ulf Molau, Jens Mogens Olesen and Hans Tybjerg and two anonymous reviewers for their constructive comments on earlier versions of this paper. We are indebted to Marianne Philipp and Per Nygaard for supervision and encouragement throughout the project. The study was funded in part by grants to both authors by The University of Copenhagen, and a Carlsberg Scholarship to IRR.

References

Beattie A (1978) Plant-animal interactions affecting gene flow in Viola. In: AJ Richards (ed) The pollination of flowers by insects, Academic Press, London, pp 151-164

Brodsgaard B, Rasmussen IR (1990) Gene flow in the species Lotus eorniculatus L. M.Sc. thesis, University of Copenhagen, Den- mark

B6cher TW (1952) Vegetationsudvikling i forhold til marin akku- mulation. I. Korshage ved indlobet tii Isefjord. English sum- mary: Plant succession in relation to marine accumulation. I. The point Korshage at the entrance to Isefjord (Seeland). Bot Tidsskr 49 : 1-32

Campbell DR, Waser NM (1989) Variation in pollen fow within and among populations of Ipomopsis aggregata. Evolution 43:1444-1455

Dreisig H (1985) Movement patterns of a clear-wing hawkmoth, Hemaris fuciformis, foraging at red catchfly, Viscaria vulgaris. Oecologia 67 : 360-366

Ellstrand NC, Devlin B, Marshall DL (1989) Gene flow by pollen into small populations: Data from experimental and natural stands of wild radish. Proc Natl Acad Sci USA 86:9044-9047

Galen C, Plowright RC (1985) The effect of nectar level and flower development on pollen carryover in inflorescences of fireweed (Epilobium angustifolium) (Onagraceae). Can Bot 63:488-491

Handel SN (1982) Dynamics of gene flow in an experimental popu- lation of Cucumis melo (Cucurbitaceae). Am J Bot 69:1538-1546

283

Heinrich B (1979) Resource heterogeneity and patterns of move- ment in foraging bumblebees. Oecologia 40:235-245

Helentjaris T, King G, Slocum M, Siedenstrand C, Wegman S (1985) Restriction fragment polymorphisms as probes for plant diversity and their development as tools for applied plant breed- ing. Plant Mol Biol 5:109-118

Hodges CM, Miller RB (1981) Pollinator flight directionality and the assessment of pollen returns. Oecologia 50:376-379

Jofuku KD, Goldberg RB (1988) Analysis of plant gene structure. In: Shaw CH (ed) Plant Molecular Biology - A practical ap- proach. IRL Press, Oxford, pp 37-51

Jones DA, Turkington R (1986) Biological flora of the British Isles. Lotus corniculatus L. J Ecol 74:1185-1212

Learn GH, Schaal BA (1987) Population subdivision for ribosomal DNA repeat variants in Clematis fremontii. Evolution 41:433-438

Levin DA (1978) Pollinator behaviour and the breeding structure of plant populations. In: Richards AJ (ed) The pollination of flowers by insects. Academic Press, London, pp 133 150

Levin DA (1981) Dispersal versus gene flow in plants. Ann Mo Bot Gard 68 : 233-253

Levin DA (1983) Plant parentage: An alternate view of the breeding structure of populations. In: King CE, Dawson PS (eds) Popu- lation biology - Retrospect and prospect. Columbia University Press, New York, pp 171-188

Levin DA (1984) Immigration in plants: an exercise in the subjunc- tive. In : Dirzo R, Sarukhan J (eds), Perspectives on plant popu- lation ecology. Sinauer, Sunderland, MA, pp 242-260

Levin DA (1988) Local differentiation and the breeding structure of plant populations. In: Gottlieb LD, Jain SK (eds), Plant evolutionary biology. Chapman and Hall, London, pp 305-329

Levin DA, Kerster HW (1968) Local gene dispersal in Phlox. Evolu- tion 22:130-139

Levin DA, Kerster HW (1969a) Density-dependent gene dispersal in Liatris. Am Nat 103:61-74

Levin DA, Kerster HW (1969b) The dependence of bee-mediated pollen and gene dispersal upon plant density. Evolution 23 : 560-571

Levin D, Kerster HW, Niedzlek M (1971) Pollinator flight direc- tionality and its effect on pollen flow. Evolution 25:113-118

Mead DA, Szczesna-Skorupa E, Kemper B (1986) Single-stranded DNA "blue" T7 promotor plasmids: a versatile tandem pro- moter system for cloning and protein engineering. Protein Engi- neering 1(1): 67-74

Milberg P (1990) Hur lfinge kan et fr6 leva? English summary: What is the maximum longevity of seeds? Sven Bot Tidskr 84: 323-352

Nevo E, Belles A, Kaplan D, Golenberg EM, Olsvig-Whittaker L, Naveh Z (1986) Natural selection of allozyme polymorphisms: A microsite test revealing ecological genetic differentiation in wild barley. Evolution 40:13-20

Nybom H, Schaal BA (1990) DNA "fingerprints" reveal genotype distributions in natural populations of blackberries and rasp- berries (Rubus, Rosaceae). Am J Bot 77:883-888

Olesen JM, Warncke E (1989) Temporal changes in pollen flow and neighbourhood structure in a population of Saxifraga hirculus L. Oecologia 79: 205-211

Plowright RC, Galen C (1985) Landmarks or obstacles: the effect of spatial heterogeneity on bumble bee foraging behaviour. Oikos 44: 459-464

Pyke G (1978) Optimal foraging: Movement patterns of bumble- bees between inflorescences. Theor Popul Biol 13:72-98

Rogers SO, Bendich AJ (1988) Extraction of DNA from plant tissues: Plant molecular biology manual A6:1-10 Kluwer Ac- ademic Publishers, Dordrecht

Sambrook J, Fritsch EF, Maniatis T (1989) Molecular cloning - A Laboratory Manual. Cold Spring Harbor Laboratory Press, New York

Schaal BA (1978) Density dependent foraging on Liatris pycnos- tachya. Evolution 32:452454

Schaal BA (1980) Measurement of gene flow in Lupinus texensis. Nature 284: 450-451

Schaal BA, Leverich WJ, Nieto-Soleto J (1987) Ribosomal DNA variation in the native plant Phlox divarieata. Mol Biol Evol 4:611-621

Schmitt J (1980) Pollinator foraging behaviour and gene dispersal in Seneeio (Compositae). Evolution 34:934-943

Schmitt J (1983) Density-dependent pollinator foraging, flowering phenology, and temporal pollen dispersal patterns in Linanthus bicolor. Evolution 37:1247-1257

Sih A, Baltus M-S (1987) Patch size, pollinator behaviour, and pollinator limitation in catnip. Ecology 68:1679-1690

Slatkin M (1987) Gene flow and the geographic structure of natural populations. Science 236:787-792

Smyth CA, Hamrick JL (1987) Realized gene flow via pollen in artificial populations of musk thistle, Carduus nutans L. Evolu- tion 41:613-619

Sokal RS, Rohlf FJ (1981) Biometry, 2nd edn, W.H. Freeman and Company, New York

Sowig P (1989) Effect of flowering plants' patch size on species composition of pollinator communities, foraging strategies, and resource partitioning in bumblebees (Hymenoptera: Apidae). Oecologia 78:550-558

Stamp NE, Lucas JR (1983) Ecological correlates of explosive seed dispersal. Oecologia 59 : 272-278

Svensson L (1985) An estimate of pollen carryover by ants in a natural population of Scleranthus perennis L. (Caryophylla- ceae). Oecologia 66:373-377

Thomson JD, Plowright RC (1980) Pollen carryover, nectar re- wards, and pollinator behaviour with spec!al reference to Dier- villa lonieera. Oecologia 46:68-74

Thomson JD, Thomson BA (1989) Dispersal of Erythronium grandiflorum pollen by bumblebees: Implications for gene flow and reproductive success. Evolution 43:657-661

Waddington KD (1981) Factors influencing pollen flow in bumble- bee-pollinated Delphinium virescens. Oikos 37 : 153-159

Waser NM, Price MV (1983) Optimal and actual outcrossing in plants, and the nature of plant-pollinator interaction. In: Jones CE, Little RL (eds) Handbook of Experimental Pollination Biology, Van Nostrand Reinhold Company Inc., New York, pp 341-359

Waser NM, Vickery RK, Price MV (1982) Patterns of seed dispersal and population differentiation in Mimulus guttatus. Evolution 36:753 761

Woodell SRJ (1978) Directionality in bumblebees in relation to environmental factors. In: Richards AJ (ed) The pollination of flowers by insects. Academic Press, London, pp 31-39

Zimmerman M (1982) Optimal foraging: Random movement by pollen collecting bumblebees_ Oecologia 53:394-398