Blackwell Science, LtdOxford, UKPSBPlant Species Biology1442-1984The Society for the Study of Species Biology, 2003December 2003182&3107121Original ArticlePOP-

ULATION SPATIAL EFFECTS ON SEED SETA. WATANABE

ET AL.

Plant Species Biology (2003) 18, 107–121

Effects of population spatial structure on the quantity and

quality of seeds set by Primula sieboldii (Primulaceae)
ATSUKO WATANABE,* KOUICHI GOKA† and IZUMI WASHITANI*
*Laboratory of Conservation Ecology, Department of Ecosystem Studies, Institute of Agricultural and Life Sciences, The University
of Tokyo, Tokyo 113-8657, Japan, ‡National Institute for Environmental Studies, Tsukuba, Ibaraki 305-8506, Japan

Abstract

Effects of spatial population structure on the reproductive success of Primula sieboldii E.

Morren, an endangered heterostylous perennial, were studied through the analysis of
quantity and quality of seeds in a natural habitat in central Japan. Seed set of individual
flowering ramets varied significantly depending on the degrees of isolation from their
potential mates. The flowering ramets, which have the opposite-morph ramets within
5 m, set significantly higher mean seeds per flower (approx. 15) than that of more isolated
ramets (approx. 5). The former produced a considerably variable number (0–100) of seeds
per flower, while the latter set more or less uniform number of seeds per flower. Viability
of seeds from the former ramets was also variable with significantly higher mean viability
percentage than the latter. Early seedling survival was also significantly different between
the maternal isolation classes, while survival and growth of the established seedlings in
a greenhouse did not differ significantly between these classes. The level of allozyme
heterozygosity of the surviving progenies was not significantly different between the
classes differing in maternal isolation. The observation suggests that the isolated mater-
nal plants produced fewer and less viable seeds, which may be a result of strong inbreed-
ing depression that manifests in early life stages of the progenies. By the stage of seedling
establishment, therefore, less fit homozygotes are purged while fitter heterozygotes are
allowed to survive irrespective of maternal isolation.
Keywords: germinability, Primula sieboldii, seed set, seedling survival, spatial population
structure.
Received 26 November 2002; revision received 18 August 2003; accepted 9 September 2003

Introduction The gene dispersal of animal-pollinated plants is

largely determined by pollinator behaviors, except for
The quality and quantity of seeds of outcrossing plants
seed dispersal. Plants in smaller populations or sparse
may differ considerably depending on the mating pattern
patches are likely to be less attractive to animal pollinators
of local populations. In populations of animal-pollinated
than large populations (Sih & Baltus 1987; Sowig 1989;
plant species, both frequency of pollinator visitation and
Ågren 1996). Lower frequency of pollinator visitation may
abundance of compatible flowers in the neighborhood
limit the deposition of compatible pollen on stigmas in
largely determine the opportunity and partner for mating
small and sparse patches. Pollen limitation due to limited
of individual plants (Handel 1983; Goodell et al. 1997).
pollinator availability often results in fertilization failure
Local population size and spatial relationship of inter-
in the short-term and breakdown of the present mating
compatible flowers may thus give rise to spatially variable
system of the plant in the long-term (Baker 1966;
breeding success and progeny genetic features if gene
Washitani 1996; Nishihiro & Washitani 1998; Matsumura
dispersal is restricted, as documented in a number of
& Washitani 2000; Cane & Tepedino 2001; Thomson 2001).
studies (Waser & Price 1983; Levin 1989; Fenster 1991;
In addition, restricted gene dispersal may contribute to
Campbell & Dooley 1992).
the establishment of kinship structure in which neighbor-
Correspondence: Atsuko Watanabe ing plants are genetically related (Turner et al. 1982; Levin
E-mail: [email protected] 1984). If proximal plants are also genetically close, the

© 2003 The Society for the Study of Species Biology

108 A . WATA N A B E E T A L .

expected consequence is increased inbreeding, followed queen bumblebee, facilitates intermorph pollination that
by heterozygote deficiency and reduced mating success result in higher seed production (Washitani et al. 1994b).
due to inbreeding depression in the neighborhood Studies on the foraging behavior of the bumblebee have
(Wright 1922; Falconer 1989; Barrett & Kohn 1991). More- shown that the interfloral movements often occur within
over, some plants may express the plasticity of mating the nearest neighborhood, with the most frequent flight
systems (Silvertown & Lovett Doust 1993); that is, the distance being less than 1 m (Waser & Price 1983; Broyles
ratio of selfing and outcrossing changes with the avail- & Wyatt 1991). Additionally, an experimental examina-
ability of compatible pollen on the maternal stigma tion of pollen transfer pattern in the Primula–Bumblebee
(Washitani 1996), which may strongly depend on the local system revealed that, in a population where multiple gen-
spatial pattern of the flowers of conspecifics. Therefore, ets were available as pollen donors, a mixed pollination
spatial structure should profoundly affect both the was brought about on a stigma by a single visit of the bee,
amount and fitness of progeny produced by individual due to the extensive pollen carryover (Matsumura &
maternal plants. Washitani 2002).
Several studies have reported positive correlations Taking advantage of this given knowledge, the present
between population size and female fecundity (Menges study attempted to test a hypothesis that the fine scale
1991; Lamont et al. 1993; Fritz & Nilsson 1994; Ågren 1996; spatial population structure influence the quantity and
see Ellstrand & Elam 1993 for reviews). The effect of inter- quality of seeds set by individual ramets of P. sieboidii in
plant distance on the mating success of self-incompatible a field population. The quantity and quality of seeds were
plants has been analyzed by several authors with hand evaluated through examining the following aspects of the
pollination experiments (Levin 1984; Sobrevila 1988; Fen- plant life stages: (1) seed set per flower; (2) seed viability;
ster 1991; Morán-Palma & Snow 1997). However, further (3) seedling survival; and (4) early growth of progeny
studies in natural systems are required for qualitative and seedlings. In addition, allozyme heterozygosity of the
quantitative analysis of the effects of spatial population progeny, which survived to the stage of seedling estab-
structure on reproductive success of plants (Sobrevila lishment, was compared in terms of the degree of spatial
1988; Heywood 1991; Kunin 1997 for reviews). isolation of maternal plants, in order to examine the pos-
The present study documents the effects of spatial iso- sibility of differential heterozygosity expression in the off-
lation of maternal plants within a local population on the spring population caused by the uneven spatial isolation
quantity and quality of the seeds of Primula sieboldii E. of the former generation.
Morren (Primulaceae) in a natural habitat. Generally, seed
production in natural habitats is affected by various biotic
and abiotic factors, among which, as is often the case, it Materials and methods
is difficult to distinguish the focal factor from others. In
Species and study site
this regard, several features of P. sieboldii make it an
appropriate species to study the effects of population spa- Primula sieboldii is an herbaceous perennial that once was
tial structure on plant reproductive success. common in a range of moist habitats in Japan. Due to
First, the distylous P. sieboldii population consists of two habitat destruction by human activities, however, the spe-
floral morphs, long- and short-styled morphs having stig- cies is rapidly decreasing in number, and size of popula-
mas and anthers differing in height, which are visually tions and is currently listed in the national red list as
distinct. Second, individual genets are identifiable within vulnerable (Environment Agency of Japan 2000). Like
a dense patch of conspecifics due to the large variation of many other Primula species, two distinct floral types, the
floral morphologies (i.e. size, color, and shape) between long- and short-styled morphs are found within popula-
genets (Washitani et al. 1994a). It needs to be noted that tions. As is typical of heterostylous species, insect pollina-
not all the conflowering plants can serve as potential mat- tors are required for legitimate pollination between the
ing partners in case of self-incompatible species. In this morphs. Previous studies have shown that queen bum-
regard, these traits of the species are advantageous to blebees (especially Bombus diversus Smith, Apidae) are
estimate potential mating partners for individual plants, the most important pollinators to facilitate intermorph
which can be readily recognized from floral morphology. pollination (Washitani et al. 1994b). Bumblebees leave
Furthermore, biotic factors affecting the reproductive suc- clear claw marks on flower petals, which can be used as
cess of the species (i.e. herbivore, pathogen and effective an index of pollinator availability. As for antagonistic
pollinator) are relatively well understood through previ- biotic agents, a flower consumer, a rove beetle, Eus-
ous studies. For instance, a specific smut fungal pathogen phaleum usphaleum bosatsu Watanabe, and a specific smut
and a floral consumer have a negative effect on female fungal pathogen, Urocrystis tranzschelina (Lavrov) Zundel
fertility of P. sieboldii (Washitani et al. 1994a; Matsumura (Ustilaginales) have been shown to lower the female
& Washitani 2000), while its predominant pollinator, the fertility of P. sieboldii (Washitani et al. 1996).

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 109

The study was conducted on a field population of P. Measurements

sieboldii in the University Forest of the University of
In mid-June of 1996 and late May of 1997, both in the late
Tsukuba (79.47 ha) located on the flank of Mt. Yatsugatake
flowering season of P. sieboldii when most flowers had
(35∞57¢-N, 138∞28¢-E, 1350 m above sea level) in Nagano
already opened, all the inflorescences of the study popu-
prefecture, central Japan. Primula sieboldii inhabits the
lation were marked, numbered and mapped.
moist forest floor dominated by Quercus mongolica spp.
An individual flowering ramet of P. sieboldii is physio-
crispula (Blume) Menitsky. The spatial population struc-
logically independent from the other ramets of the same
ture of P. sieboldii in this habitat was a typical meta-pop-
genet. The pedicels of flowering ramets were marked with
ulation, characterized by the hierarchical composition of
small strips of vinyl chloride tape. Numbers of flowering
the population subunits defined as follows: ‘a patch’ was
ramets per genet and flower number per ramet were
defined as an aggregation of flowering ramets separated
recorded for each patch. Floral morph type (long- or short-
from their closest conspecifics by > 15 m, while ‘a subpop-
styled), the presence or absence of consumer insects
ulation’ defined as an aggregation of patches separated
(mainly rove beetles), presence of infection by smut fun-
by > 50 m. In this way, the major part of observed ramets
gus and bumblebees’ claw marks on flower petals were
were categorized into four subpopulations (SpA, SpB,
recorded for each ramet (Washitani et al. 1996). Visible
SpC, SpD), inclusive of 31 patches (Fig. 1). Among these
traces of herbivory on any part of the corolla were
subpopulations, large spatial variation in the amount of
regarded as damage caused by the rove beetles (Washitani
interaction between P. sieboldii and its pathogens and
et al. 1996). The distance to the nearest ramet of the com-
herbivores have been observed (Washitani et al. 1996).
patible (opposite) morph was measured for individual
flowering ramets. The intervals more than 5 m were paced
out, while the shorter ones measured by the centimeter.

Seed set
From late July to early August in 1996 and 1997, matured
capsules were collected from the marked ramets. The
number of mature capsules per ramet and the number of
seeds in each capsule were counted and the presence or
absence of herbivory and pathogenic infection were
recorded. Mean female fertility (seed set per flower) was
calculated for individual flowering ramets. Genet and
subpopulation means for female fertility were calculated
from the above data.

Seed viability
Seed viability was assessed by germinability after full
removal of primary dormancy of the seeds. An effective
condition for breaking the dormancy of P. sieboldii seeds
is a combination of moist-chilling and temperature fluc-
tuation (Washitani & Kabaya 1988). Mature seeds col-
lected in 1996 were stored at 4∞C in a dry state until used
for the experiment. After 8 weeks of moist chilling, 25
seeds from an individual maternal ramet were transferred
to 3-cm diameter Petri dishes containing a double layer
of filter paper, wet with distilled water. Four replications
were made for each maternal genet (n = 89). The seeds
were incubated under a photoperiod of 12/12 h dark/
Fig. 1 The distribution of Primula sieboldii flowering ramets in
light and an alternation of temperature with 12/25∞C.
the University Forest location, on the flank of Mt. Yatsugatake.
Germination was determined by a visible radicle emer-
The shaded circle, , indicates the location and size (number of
flowering ramets) of a patch. A patch is defined as an aggregation gence, which was inspected daily during the incubation.
of flowering ramets separated from their closest conspecifics After 30 days of incubation, ungerminated seeds were
by > 15 m. incubated for 18 h in an aquatic solution of 100 mM gib-

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
110 A . WATA N A B E E T A L .

berellin (GA3 in soluble granule; Kyowa Co., Tokyo, ln W2 - ln W1

Japan) for the breakage of remaining dormancy (Bread- RGR =
14
( g◊g -1◊d -1 )
beer 1988). Newly germinated seeds were counted like-
wise. The cumulative germination percentage, until the where W1 and W2 refer to the estimated biomass at pre-
day after which no additional germination had observed, ceding and following measurements, respectively. The
was defined as seed viability. In 1997, the viability of the measurement of RGR was performed for 1996 seedlings.
seeds was tested with the same procedure as in 1996,
except the seeds were treated with GA3 immediately after
Electrophoresis and enzyme staining
4 weeks of moist chilling, since it was proved in the pre-
vious year that the GA3 treatment was effective in break- Newly expanded leaves were sampled from the seedlings
ing seed dormancy of P. sieboldii. derived from the seeds set in 1996, which had survived to
the last stage of the growth experiment in the greenhouse.
Heterozygosity of the seedlings from four patches,
Seedling survival and early growth
which differs in the degree of spatial isolation of their
Seedlings obtained in the above experiment were planted maternal plants in their original habitat, were compared.
to moist peat moss plates (20 ¥ 12 ¥ 3 cm; Sakatanotane The patches of maternal ramets that had the compatible
Co., Yokohama) immediately after germination. Within a morph within 5 m were categorized as ‘unisolated patch’.
plate, a grid with 50 cells was predetermined in a 5 ¥ 10 The maternal plants of the seedlings in the ‘isolated patch’
coordinate system. The size of each cell was large enough were separated from its nearest ramet of opposite morph
to avoid possible interactions among seedlings. A newly by more than 5 m. Leaf samples from 36, 14 and eight
germinated seedling was assigned with a three-digits ran- seedlings (derived from 12, seven and four maternal gen-
dom number, which represents the plate number, x- and ets, respectively) were corrected from three unisolated
y- coordinates of the plate, respectively. In this method, patches, and eight seedlings (from four maternal genets)
the seedlings were arranged in a randomized order to from an isolated patch (Table 6).
avoid possible location effects within/among the plate(s). Leaf materials were stored at -80∞C until electrophore-
Total of 20 plates were placed in lines onto benches in a sis. A sample of 100 mg leaf material was homogenized
greenhouse maintained at 25∞C day/17∞C night tempera- in 100 mL of an extraction buffer containing 93 mM Tris-
tures and rotationally repositioned once every 3 days. The HCl (pH 7.5), 23.4% glycerol, 0.6% Triton-X100, 2.3 mM
seedlings were grown under natural light regime until the DTT, 2.8 mM EDTA, 0.6 mM bNAD+, 0.5 mM bNADP+
first leaf expansion. The number of seedlings, which did and 0.05% 2-mercaptoethanol. Polyacrylamide vertical
not produce any leaf and/or died on the plates, was slab gel electrophoresis and enzyme staining were per-
recorded as seedling mortality. After the first true leaf formed to screen loci on the following enzymes (in paren-
expanded (the stage was defined as seedling establish- theses, abbreviation, E.C. number and reference): esterase
ment), seedlings were replanted individually in unglazed (EST, 3.1.1, Goka & Takafuji 1992, 1995), fumarase (FUM,
pots (12 cm diameter, 7 cm in depth) containing potting 4.2.1.2, Brewer 1970), malate dehydrogenase (MDH,
compost, and grown in the same greenhouse. Pots were 1.1.1.37, Goka & Takafuji 1992, 1995), maric enzyme (ME,
placed under mesh screen (opening size of strand: 1.1.1.40, Soltis et al. 1983), phosphoglucoisomerase (PGI,
0.6 mm ¥ 0.95 mm, light transmission ± SD = 0.83 ± 0.02 5.3.1.9, Goka & Takafuji 1992, 1995), phosphoglucomutase
[n = 28]). Plants were watered once a day, or twice during (PGM, 5.4.2.2, Tanksley 1979), 6-phosphogluconate dehy-
the mid summer days. Liquid fertilizer (¥ 1000 Hyponex drogenase (6PGDH, 1.1.1.44, Soltis et al. 1983), shikimic
containing 5% N, 10% P, and 5% K; Hyponex Japan Co., dehydrogenase (SKDH, 1.1.1.25, Soltis et al. 1983). Sample
Osaka) was applied every 10 days. extraction and electrophoresis were carried out between
After a week of seedling establishment, biomass of each 0∞C and 4∞C following the protocol of Goka & Takafuji
seedling was estimated in a non-destructive manner from (1992).
the product of the total number of leaves per plant (coty-
ledons not included), the length and the width of the
Statistical analyzes
largest leaf present at the time of measurements. Linear
regression of dry weight and the estimate of size were The relationship between subpopulation and yearly
highly significant in a preliminary experiment (r2 = 0.89, occurrences of antagonistic biological interactions were
d.f. = 50, P < 0.0001). The size parameters for biomass esti- tested by chi-square test of R ¥ C contingency tables.
mation were measured every 2 weeks for a 12-week of The following aspects were examined as parameters for
growth period after transplantation. The relative growth spatial population structure of P. sieboidii: (1) the number
rate (RGR) for each period between two successive mea- of all flowering ramets within a patch; (2) the number of
surement days was calculated as: the opposite morph ramets within a patch; (3) the number

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 111

of the opposite morph genets within a patch; and (4) the observation years. Although pollinator availability
distance to the nearest flowering ramet of the opposite indexed by bumblebee’s claw marks was relatively high
morph. The relationship between the above four param- over the entire study site (63.96% of all ramets), spatial
eters and quantity or quality of seeds (genet mean female variation in the pollinator visitation among the subpopu-
fertility, seed viability, seedling survival by establishment, lations was significant (c2 = 17.22, P = 0.0006 in 1996,
and the maximum relative growth rate [RGRmax] of seed- c2 = 153.36, P < 0.0001 in 1997). The claw mark occurrence
ling) were analyzed on genet mean by correlation test. was constantly higher in SpB and SpC than in SpA and
The analysis of covariance (ANCOVA) was employed to SpD (Fig. 2).
compare the effects of morph and each of the four param- The female fertility was significantly different among
eters of spatial population structure on seed set, germina- subpopulations but not the flower morphs (ANOVA,
tion performance, seedling survival by establishment and Table 1)
on the maximum relative growth rate (RGRmax) of progeny In SpA and SpB, relatively high percentages of flowers
seedlings. The final percentages of germination and seed- bore signs of herbivory by Eusphareum beetles (Fig. 2) and
ling survival were arcsine transformed prior to analysis. female fertility was consistently low during the 2 years.
Paired comparisons for the means were made by Fisher’s In SpD, a fraction of flowers were infected by smut fungus
protected least significant difference (PSLD). and the seed set per flower was lower than that of SpC.
In order to examine the hypothesis that the rate of In SpC, in which only a negligible fraction of ramets were
inbreeding is higher in isolated plants than in unisolated attacked by rove beetle and virtually no infection was
plants, the values of heterozygosity were compared detected for both observation years (Fig. 2), female fertil-
between the progenies from maternal plants from four ity was significantly higher than other three subpopula-
patches differing in the distance to the potential mate. tions for both years (Fisher’s PSLD test: P < 0.0001 for
Specifically, the proportion of polymorphic loci and each pair with SpC).
observed and expected heterozygosity (Hobs and Hexp)
were calculated from the allozyme data. Deviations from
Seed set
Hardy–Weinberg expectations were examined for each
polymorphic locus within each patch by calculating In the present study, SpC, in which the impacts of antag-
Wright’s fixation index, Fis (= 1 - Hobs/Hexp) (Wright 1922). onistic biological interactions were negligible (Fig. 2), was
Fixation indices were tested for significance (F should chosen for the analysis of the relationship between seed
equal zero under panmixia in the absence of selection) by set and spatial pattern of flowering ramet (Table 2). The
c2 tests (Li & Horvitz 1953). Statistical significance of the number of flowering ramet per patch (rs = 0.396, P < 0.01),
difference in Hobs between patches was tested by c2 test. the number of flowering ramet of the opposite morph
Statistical analysis was done using StatView version 4.5 (rs = 0.369, P < 0.05) and the number of flowering genet of
(SAS Institute Inc., Cary, NC, USA) and Super ANOVA the opposite morph (rs = 0.335, P < 0.01) showed signifi-
version 1.11 (Abacus Concepts, Inc., Berkeley, CA). cant correlation with genet mean seed set per flower in
1996, but neither showed significant correlation in 1997.
In both years, a significant negative correlation was
Results detected between genet mean seed set per flower and the
distance to the nearest flowering ramet of the opposite
Population spatial structure and biological interaction
morph (rs = - 0.24, P < 0.0001 in 1996, rs = - 0.31,
The spatial population structure of P. sieboldii was charac- P < 0.0001 in 1997). The results of ANCOVA showed the
terized by the hierarchical composition of the population distance had a significant effect on the seed set for two
subunits of subpopulations and patches, as described in years, but morph only the first year (Table 2). Seed set of
‘Species and study site’ section. ramets that had a flowering ramet of the opposite morph
The proportion of infected flowers was significantly within 5 m varied considerably (0–100 seeds per flower),
related to subpopulations over the two study years whereas less variable and fewer seeds were set by more
(c2 = 110.67, P < 0.0001 for 1996, c2 = 14.77, P = 0.002 for isolated ramets (Fig. 3), that is, the distance and seed
1997). The post-hoc test showed that SpA had the largest quantity were not in linear relationship.
damage, with only SpC being free from fungal infection Based on this observation and an existing report of
for the 2 years. neighborhood area (Nishihiro et al. 2000), 5 m was
Differences in the proportion of insect herbivory were adopted as a threshold of population spatial structure that
also significant among the subpopulations (c2 = 142.73, may affect the female reproductive performance. In order
P < 0.0001 for 1996, c2 = 322.19, P < 0.0001 for 1997). No to examine the relationship between the distance to
insect damage was observed in SpC or SpD, while a frac- potential mate and the seed set, maternal ramets were
tion of flowers were consumed in SpA and SpB for both classified into two classes according to the distance to the

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
112 A . WATA N A B E E T A L .

Table 1 Results of ANOVA of the effects of subpopulation and

morph on seed set per flower of Primula sieboldii in the four
subpopulations of the University Forest. F-value is shown with
its significance level

Year Source of variance d.f. MS F-value

1996 Subpopulation 3 9050.89 21.37***

Morph 1 77.61 0.55
Subpop. ¥ Morph 3 986.23 0.73
Residual 671 94706.59
1997 Subpopulation 3 26515.98 53.41***
Morph 1 18.30 0.11
Subpop. ¥ Morph 3 7.79 0.05
Residual 856 165.50

*** P < 0.001

nearest flowering ramet of the opposite morph: flowering

ramets, which have the potential mate within a 5-m radius
(unisolated class), and more isolated flowering ramets
(isolated class). Difference in seed set between the isola-
tion classes was highly significant (ANOVA, d.f. = 1,
F = 11.8, P < 0.0001 in 1996, d.f. = 1, F = 29.5, P < 0.0001 in
1997), with flowers of unisolated class producing more
seeds (mean seed set per flower ±1 SD: 13.5 ± 1.0 in 1996,
16.7 ± 0.9 in 1997) than flowers of isolated class (6.28 ± 1.5
in 1996, 5.41 ± 1.0 in 1997) regardless of morph for both
years.
More flowering ramets in unisolated classes (76%) had
been visited by queen bumble bees than those in isolated
classes (59%). The difference in the proportion of claw
marked flowers between the two classes was significant
(c2 = 15.70, P < 0.0001)

Seed viability
Seed viability was highly variable, depending on mater-
nal genet, and ranged from 0 to 73% in 1996, and 0 to 98%
in 1997. The GA3 treatment prior to temperature alternat-
ing incubation in 1997 resulted in higher mean final per-
centage of germination, but the overall tendency of
germinability in relation to population size and spatial
structure was consistent for both experiment years
(Fig. 3).
For the experiment performed in 1996, significant pos-
itive correlations were found between patch size (rs = 0.50,
P = 0.0004), the number of the opposite morph ramets per
patch (rs = 0.43, P = 0.0005), and between the number of
the opposite morph genets per patch (rs = 0.50, P = 0.0005)
and seed viability. Although there was a tendency of
Fig. 2 Between subpopulations of (a) flower consumption by
increasing viability with each of the three patch size
Eusphhareum beetles, (b) infection by smut fungi, Urocrystis tran-
zschelina, (c) flower visitation by Bombus diversus queens and (d) parameters, correlations were not statistically significant
genet mean seed set per flower. Observation made in 1996 (
) for the 1997 data. However, there was a consistent and
and 1997 (), n = sample size. significant negative relationship between distance to the

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 113

Table 2 Results of ANCOVA of the effects of

parameters of spatial structure population Year Source of variance d.f. MS F-value
and morph on seed set per flower of Prim-
1996 Distance to the nearest ramet of the opposite morph 1 963.66 5.397*
ula sieboldii in the studied subpopulation
Morph 1 1050.22 5.881*
(SpC) of the University Forest. F-value is
Distance ¥ Morph 1 377.54 2.114
shown with its significance level
Residual 233 41605.66
Patch size 1 1546.90 8.804**
Morph 1 51.13 0.291
Patch size ¥ Morph 1 109.23 0.622
Residual 233 40939.67
Number of opposite morph ramet 1 1539.93 8.744**
Morph 1 77.70 0.441
Number of opposite morph ramet ¥ Morph 1 238.28 1.353
Residual 233 41032.17
Number of opposite morph genet 1 49.97 9.044**
Morph 1 1590.70 0.284
Number of opposite morph genet ¥ Morph 1 317.13 1.803
Residual 233 40982.73
1997 Distance to the nearest ramet of the opposite morph 1 9554.70 41.061***
Morph 1 151.74 0.652
Distance ¥ Morph 1 703.08 3.021
Residual 456 106108.50
Patch size 1 2039.28 8.184**
Morph 1 3.70 0.015
Patch size ¥ Morph 1 0.45 0.002
Residual 456 113624.76
Number of opposite morph ramet 1 2126.34 8.542**
Morph 1 2.78 0.011
Number of opposite morph ramet ¥ Morph 1 1.11 0.004
Residual 456 113507.48
Number of opposite morph genet 1 2225.68 8.944**
Morph 1 0.05 0.000
Number of opposite morph genet ¥ Morph 1 55.00 0.221
Residual 456 248.85

*P < 0.05, **P < 0.01, *** P < 0.001

nearest ramet of the opposite morph and seed viability for did not survive until the first true leaf expanded. In 1996,
both years (rs = - 0.33, P = 0.0004 in 1996, rs = - 0.48, patch size (rs = 0.46, P = 0.0006), number of opposite
P < 0.0001 in 1997). ANCOVA results also showed the sig- morph ramets (rs = 0.41, P = 0.0025), number of opposite
nificant effect of the distance parameter on seed viability morph genets (rs = 0.48, P = 0.0003) and the distance to
for 2 years (Table 3). the nearest ramet of the opposite morph (rs = - 0.33,
When analyzed by isolation classes of maternal plants, P = 0.0019), showed significant correlation with the per-
seed viability was lower for seeds produced by maternal centage of seedling survival. In 1997, however, the corre-
plants of the isolated class (mean germination percentage lations were not significant for the three patch size
±1 SD: 5.67 ± 2.13 in 1996, 39.14 ± 7.51 in 1997), compared parameters, but the distance parameter (rs = - 0.26,
to those of unisolated class did (21.36 ± 2.73 in 1996, P = 0.0013). The analysis of covariance showed the signif-
61.34 ± 4.14 in 1997). The flower morph did not cause icant effect of the distance parameter on the seedling sur-
significant difference in seed viability, but the distance vival for 2 years (Table 4). When analyzed by isolation
class (ANOVA, d.f. = 1, F = 9.45, P = 0.003 in 1996, d.f. = 1, classes, early seedling mortality before establishment
F = 5.96, P = 0.018 in 1997). was significantly higher (F = 4.82, P < 0.0001 in 1996,
F = 2.19, P = 0.03 in 1997) for the progenies of isolated
plants (90% in 1996, 81% in 1997) than ones of unisolated
Seedling survival and early growth
plants (75% in 1996, 68% in 1997). Once established, how-
Seedling mortality was considerably high during the first ever, almost all the remaining seedlings, irrespective of
2 weeks after germination and a large proportion maternal isolation class, survived for 12 weeks after
(approximately 88% in 1996, 52% in 1997) of seedlings transplantation.

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
114 A . WATA N A B E E T A L .

Relative growth rate of the seedlings, which survived

to the stage of first true leaf expansion, was highest for
the first 2 weeks after the true leaf expansion. Then RGR
decreased as plants grew, as commonly observed in all the
surviving seedlings regardless of floral morph (Fig. 4). No
significant effect of morph or spatial population structure
was detected on RGRmax of the survived seedlings
(ANCOVA, Table 5).

Heterozygosity of progeny seedlings

One locus each for three enzymes (EST, SKDH, PGM)
expressed recognizable variation. Allele frequencies and
single locus allelic diversity were calculated for these
three polymorphic loci. The measures of single-locus
allelic heterozygosity are given in Table 6. Observed het-
erozygosity was significantly higher than the expected
value for some loci of unisolated patches. From the het-
erozygosity of the isolated patch, the significant deviation
from Herdy–Weinberg equilibrium was not detected
(Table 6).

Discussion
Population spatial structure and quantity of seed set
The population spatial structure of a species cannot be
independent from other environmental factors in natural
habitats, which may have a substantial effect on the seed
set and progeny fitness. In the present study, to minimize
the effect of unknown factors, the analysis was performed
in a single subpopulation (SpC), in which antagonistic
biological interactions were not crucial. In this subpopu-
lation, it was possible to find some of the factors that
influenced seed set of P. sieboldii. Specifically, patch size
and isolation of plants were shown to markedly affect
female fertility of individual flowering ramets of the spe-
cies. Through the analysis of pollinator availability, higher
pollinator visitation was found in small and isolated
patches. From these results, lower compatible pollen dep-
osition is expected for isolated plants since bumblebees
prefer aggregated patch (Klinkhamer & de Jong 1990).
Previous studies revealed that shortages in the supply of
compatible pollen, due to pollinator limitation, are
Fig. 3 Relationship between (a) genet mean seed set per flower, responsible for low female fertility of P. sieboldii popula-
(b) % viable seeds, (c) % seedling survival, (d) RGRMAX after tion of an isolated nature reserve in an urbanized land-
seedling establishment and distance to the nearest flowering scape (Washitani et al. 1994a) and of a fragmented
ramet of the opposite morph in the studied subpopulation (SpC) agricultural landscape (Matsumura & Washitani 2000).
in the University Forest population of Primula sieboldii for () Unlike these local populations, the overall pollinator
1996 and () 1997. The seed viability was assessed by the pro-
availability is thought to be fairly sufficient in the studied
portion of germinable seeds to the total seed set after breaking
population. Nevertheless, at fine scale, the present study
dormancy. Percentage seedling survival was evaluated by com-
paring the proportion of surviving seedlings to the total number demonstrated a spatial heterogeneity of pollinator visita-
of seed samples at 2 weeks after the emergence of the first true tion and of plant fertility within a subpopulation.
leaf. Observed high seed set in aggregated plants should be

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 115

Table 3 Results of ANCOVA of the effects of

parameters of spatial structure population Year Source of variance d.f. MS F-value
and morph on seed viability of Primula sie-
1996 Distance to the nearest ramet of the opposite morph 1 0.15 5.421*
boldii in the studied subpopulation (SpC)
Morph 1 0.02 0.797
of the University Forest. F-value is shown
Distance ¥ Morph 1 0.02 0.592
with its significance level
Residual 47 1.33
Patch size 1 0.34 13.184***
Morph 1 0.01 0.243
Patch size ¥ Morph 1 3.06E-3 0.097
Residual 47 1.21
Number of opposite morph ramet 1 0.21 7.642
Morph 1 0.01 0.126
Number of opposite morph ramet ¥ Morph 1 4.07E-5 0.001
Residual 47 1.30
Number of opposite morph genet 1 0.34 13.451***
Morph 1 0.01 0.506
Number of opposite morph genet ¥ Morph 1 3.26E-3 0.109
Residual 47 1.20
1997 Distance to the nearest ramet of the opposite morph 1 12.28 12.628***
Morph 1 0.17 1.636
Distance ¥ Morph 1 0.33 3.240
Residual 47 4.77
Patch size 1 0.46 3.855
Morph 1 0.25 2.036
Patch size ¥ Morph 1 0.29 2.443
Residual 47 4.21
Number of opposite morph ramet 1 0.41 3.344
Morph 1 0.31 2.493
esidual 47 4.28
Number of opposite morph genet 1 0.39 3.449
Morph 1 0.48 4.311
Number of opposite morph genet ¥ Morph 1 0.76 6.724
Residual 47 3.70

*P < 0.05, **P < 0.01, ***P < 0.001

ascribed to more frequent pollinator visitation and higher small and isolated patches. However, in this study, it was
probability to receive compatible pollen from the opposite not possible to separate the effects of these parameters
morph flowers within the neighborhood. Among the (size, density and interval of potential mates) because
parameters of spatial population structure, the patch size they are usually correlated to each other in the field pop-
should affect the attractiveness to pollinators and size of ulation. An empiric examination of population structure
effective pollen load for individual ramet. Under a con- should be required for further discussion.
certed effect of the two factors, increased pollinator (pol- The restriction in the availability of either mating part-
len) availability should bring about a greater seed set. ner or pollinator could lead to pollen limitation, therefore
Mating partner availability, which is largely deter- resulting in lower seed set in isolated flowering ramets.
mined by the degree of isolation from the compatible The causatives of the reduction in seed set may be
flowers, may be another important limiting factor for pol- explained prezygotically (e.g. small stigmatic pollen load
len transfer as much as pollinator availability (Nishihiro size, morphological incompatibility, pollen tube endur-
& Washitani 1998). Among parameters representing pop- ance, pollen tube growth rates) (Johnston 1993; Dogterom
ulation spatial structure, distance to the nearest flowering et al. 2000; Marshall et al. 2000; Marshall & Diggle 2001;
ramet of the opposite morph was suggested as an impor- Pasonen et al. 2001) or postzygotically (e.g. fruit abortion,
tant index of the mating partner availability in the present early acting inbreeding depression) (Stephanson 1981;
study. Flowering ramets belonging to large and dense Casper 1988; Becerra & Lloyd 1992; Montalvo 1992), either
patches of the species had potential mating partner in a of which needs further investigation on the post pollina-
shorter distance and set more seeds compared to those in tion mechanism for this species.

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
116 A . WATA N A B E E T A L .

Table 4 Results of ANCOVA of the effects of

Year Source of variance d.f. MS F-value parameters of spatial structure population
and morph on seedling survival of Primula
1996 Distance to the nearest ramet of the opposite morph 1 0.15 5.421*
sieboldii in the studied subpopulation
Morph 1 0.02 0.797
(SpC) of the University Forest. The seed-
Distance ¥ Morph 1 0.02 0.592
ling survival was assessed by the propor-
Residual 47 1.33
tion of seedlings surviving to the
Patch size 1 0.34 13.184***
emergence of the first true leaf to the total
Morph 1 0.01 0.243
number of sample seeds. F-value is shown
Patch size ¥ Morph 1 3.06E-3 0.097
with its significance level
Residual 47 1.21
Number of opposite morph ramet 1 0.21 7.642**
Morph 1 0.01 0.126
Number of opposite morph ramet ¥ Morph 1 4.07E-5 0.001
Residual 47 1.30
Number of opposite morph genet 1 0.34 13.451***
Morph 1 0.01 0.506
Number of opposite morph genet ¥ Morph 1 3.26E-3 0.109
Residual 47 1.20
1997 Distance to the nearest ramet of the opposite morph 1 12.28 12.628***
Morph 1 0.17 1.636
Distance ¥ Morph 1 0.33 3.240
Residual 47 4.77
Patch size 1 0.46 3.855
Morph 1 0.25 2.036
Patch size ¥ Morph 1 0.29 2.443
Residual 47 4.21
Number of opposite morph ramet 1 0.41 3.344
Morph 1 0.31 2.493
Number of opposite morph ramet ¥ Morph 1 0.39 3.176
Residual 47 4.28
Number of opposite morph genet 1 0.39 3.449
Morph 1 0.48 4.311
Number of opposite morph genet ¥ Morph 1 0.76 6.724
Residual 47 3.70

*P < 0.05, **P < 0.01, *** P < 0.001

Population spatial structure and quality of the seeds

The present study showed that the germination perfor-

mance of progeny was influenced by population spatial
structure in the natural habitat. Isolated maternal plants
produced significantly less germinable seeds compared to
those in large and dense patches. The observed differen-
tial progeny performance may be explained by the com-
bined effect of pollen competition between outcross
donors and inbreeding depression, in relation with pop-
ulation spatial structure.
The spatial aggregation of diverse genets may
facilitate mixed pollination by bumblebees, as sug-
gested by a previous study (Matsumura & Washi-
tani 2002). The mixed pollination enables multiple
Fig. 4 Courses of change of long-() and () short-style mater- paternity, which is supposed to increase reproduc-
nal morphs in the relative growth rate (RGR) of Primula sieboldii
tive success of insect-pollinated, self-incompatible
seedlings established in a greenhouse. Biomass of the seedling
plant (Marshall & Ellstrand 1986; Marshall 1990,
was estimated from the regression model: dry mass (g) =
-3.5 ¥ 10-4 + 1.14 ¥ 10-5 ¥ number of true leaves ¥ length of the 1991). This is explained as a result of pollen compe-
largest leaf (mm) ¥ width of largest leaf (mm) (r2 = 0.89). Vertical tition between outcross donors under the supply of
bars indicate ± 1 SD. genetically diverse pollen grains. The progenies pro-

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 117

Table 5 Results of ANCOVA of the effects of

parameters of spatial structure population Year Source of variance d.f. MS F-value
and morph on RGRmax of progeny seed-
1996 Distance to the nearest ramet of the opposite morph 1 0.032 6.226
lings of Primula sieboldii derived from the
Morph 1 0.011 2.172
studied subpopulation (SpC) of the Uni-
Distance ¥ Morph 1 0.012 2.309
versity Forest. All F-values were not
Residual 108 0.552
significant
Patch size 1 0.005 1.003
Morph 1 1.128E-4 0.022
Patch size ¥ Morph 1 0.006 1.247
Residual 108 0.546
Number of opposite morph ramet 1 0.003 0.524
Morph 1 0.002 0.489
Number of opposite morph ramet ¥ Morph 1 0.013 2.539
Residual 108 0.544
Number of opposite morph genet 1 0.013 2.631
Morph 1 0.001 0.195
Number of opposite morph genet ¥ Morph 1 0.001 0.253
Residual 108 0.549

Table 6 H statistics of allelic diversity at

esterase (EST), shikimic dehydrogenase Patch 1 (n = 36) Patch 2 (n = 14) Patch 3 (n = 8) Patch 4 (n = 8)
(SKDH), phosphoglucomutase (PGM) and unisolated unisolated unisolated isolated
average over the three loci for seedlings Hexp Hobs Hexp Hobs Hexp Hobs Hexp Hobs
from maternal plants belonging to ‘uniso-
EST 0.291 0.361 0.640 0.710 0.469 0.750 0.219 0.250
lated’ and ‘isolated’ patches. The statistical
SKDH 0.475 0.667 0.490 0.570 0.469 0.750 0.375 0.500
significance of the deviation from Hardy-
PGM 0.475 0.677 0.070 0.070 0.469 0.750 0.219 0.250
Weingerg is shown with its level. Wright’s
Mean 0.414 0.565 0.400 0.450 0.469 0.750 0.271 0.333
fixation index: Fis
Fis Fis Fis Fis
EST 0.194*** 0.100*** 0.375*** 0.125
SKDH 0.287 0.140 0.375 0.250
PGM 0.287** 0.040*** 0.375 0.125
Mean 0.256 0.090 0.375 0.167

**P < 0.01, ***P < 0.001

duced by better combination of parental zygotes morph of P. sieboldii capture a large amount of selfing
should have increased fitness. pollen grains (Washitani et al. 1994a; Nishihiro et al. 2000).
In this line of reasoning, the higher performance of The flowers deficient in compatible pollen deposition
progeny from the aggregated patch, as observed in the may increase selfing rate if they have partial or cryptic
present study, may be ascribed to the genetic variation of self-compatibility as suggested for this species (Washitani
pollen deposit on maternal stigma. Since the precise et al. 1991, 1994a; Nishihiro et al. 2000).
assessment of the relationship between pollination pat- Either selfing or crossing between close relatives (bipa-
tern and progeny performance is difficult to evaluate in rental inbreeding, Heywood 1993) produces offspring
the field, further discussion on progeny performance in that may suffer from inbreeding depression. A common
relation to the number of pollen donors should be based finding of studies in attempts of measuring inbreeding
on an examination through carefully designed experi- depression for herbaceous plants is that effects of inbreed-
mental pollinations. ing are stronger in late life stages (Schoen 1983; Kohn
However, in an isolated patch, not only the quantity but 1988; Barrett & Kohn 1991; for reviews). In contrast, the
the genetic diversity of pollen delivered to the stigma may present study showed that the fitness components in early
be lower than that of an aggregated patch, because of life stages (seed set per flower, seed viability and early
fewer mating partners within the range of pollinator’s survival of the seedlings) were significantly lower in the
feeding activity, thus resulting in a lower levels of pollen progenies of more isolated maternal plants. In an environ-
carryover. In the absence of pollinator visitation, as pre- ment marked by low availability of both mating partner
vious studies have suggested, the stigmas of long-styled and pollinator, the decreased seed set may be ascribed to

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
118 A . WATA N A B E E T A L .

Fig. 5 Schematic diagram for the explanation of cause–effect relationship between the variation in mating partner and pollinator
availability and the quality and quantity of seed set. The circled S and O indicate the self and the outcross pollen on the stigma,
respectively. The outcross pollen, from different genets, are shown in different patterns. (a) For a single ramet in a dense patch (left),
where multiple genets are flowering, mating partner and pollen availability are considered to be higher than that of an isolated ramet
(right). (b) The stigmas of an unisolated ramet (left) should receive more genetically diverse pollen deposits than the isolated ramet,
reflecting mating partner availability in the proximal plant community. Without pollen visitation, the stigmas may only be loaded by
self pollen (right). (c) The seed set should be larger under larger pollen load size composed of diverse paternal origins. For sparse and
isolated patch, some flowers may produce seeds from selfing. The selfed seeds may be purged before seedling establishment. If a ramet
had rigorous self incompatibility, it may not produce seeds without a supply of compatible pollen. Then the reproduction fails at this
stage. (d) The selfed and inbred progeny may be purged due to strong inbreeding depression by the seedling establishment stage. Due
to the purging effect, only a few seedlings should survive from the patches where the mating partner and pollinator availability are
moderate, the surviving few can compete with those produced in the dense patches.

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 119

the self-incompatibility of the species, and the lower seed Becerra J. X. & Lloyd D. G. (1992) Competition-dependent abscis-
viability and progeny fitness may result from inbreeding sion of self-pollinated flowers of Phormium tenax (Agvaceae):
depression. a second action of self-incompatibility at the whole flower
level? Evolution 46: 458–469.
Breadbeer J. W. (1988) The Breaking of Seed Dormancy. Blackie
Allozyme heterozygosity and purging effect of Academic & Professional, London.
inbreeding depression Brewer G. J. (1970) An Introduction to Isozyme Techniques. Aca-
demic Press, New York, pp. 186.
Theoretically, heterozygosity is predicted to be lower in
Broyles S. B. & Wyatt R. (1991) Effective pollen dispersal in a
inbreeding populations (Charlesworth & Charlesworth natural population of Asclepias exaltata: the influence of polli-
1987). In the present study, however, allozyme heterozy- nator behavior, genetic similarity, and mating success. Amer-
gosity of the established seedlings was not different from ican Naturalist 138: 1239–1249.
the value expected under Hardy–Weinberg equilibrium Campbell D. R. & Dooley J. L. (1992) The spatial scale of genetic
for those derived from the isolated patch. Besides, the differentiation in a hummingbird-pollinated plant: compari-
son with models of isolation by distance. American Naturalist
heterozygosity was rather higher for unisolated patches.
139: 735–748.
The results suggest that higher magnitude of inbreed-
Cane J. H. & Tepedino V. J. (2001) Causes and extent of declines
ing depression is manifested early in the diploid stage of among native North American invertebrate pollinators:
the life cycle of P. sieboldii and the purging effect of detection, evidence, and consequences. Conservation Ecology
inbreeding depression sweeps away the less fit homozy- 5: 1.
gotes by the stage of seedling establishment. Therefore, Casper B. B. (1988) Evidence for selective embryo abortion in
once established, the surviving few progeny of isolated Ctyptantha flava. American Naturalist 132: 318–326.
Charlesworth D. & Charlesworth B. (1987) Inbreeding depres-
maternal plants compete with those of unisolated plants
sion and its evolutionary consequences. Annual Review of
as far as heterozygosity is concerned (Fig. 5).
Ecology and Systematics 18: 237–268.
Since the expression of inbreeding can differ in magni- Dogterom M. H., Wilson M. L. & Mukai A. (2000) Effect of pollen
tude among life stages and late-acting inbreeding is sup- load size and sources (self, outcross) on seed and fruit pro-
posed to be difficult to purge (Husband & Schemske duction in highbush blueberry cv. ‘Bluecrop’ (Vaccinium
1996), further discussion about life-time progeny fitness corymbosum; Ericaceae). American Journal of Botany 87: 1584–
awaits for the examination on the extent of inbreeding 1591.
Ellstrand N. C. & Elam D. R. (1993) Population genetic conse-
depression at later life stages, which was not included in
quences of small population size: implications for plant con-
the present study.
servation. Annual Review of Ecology and Systematics 24: 217–
242.
Acknowledgements Environment Agency of Japan. (2000) Threatened Wildlife of
Japan: Red Data Book 2nd edn, , Vol. 8. Vascular Plants. Japan
The authors thank Mr J. Nishihiro (Tokyo University) for Wildlife Research Center, Tokyo.
his helpful comments of the early draft of the manuscript; Falconer D. S. (1989) Introduction to Quantitative Genetics. Long-
Miss M. Uchiyama (Tokyo Metropolitan University) for man, London.
assistance in the fieldwork; the staff of the Agricultural Fenster C. B. (1991) Gene flow in Chamaecrista fasciculata (Legu-
and Forestry Research Center and University Forest, minosae) I. Gene dispersal. Evolution 45: 398–409.
Tsukuba University. This study was supported by the Fritz A.-L. & Nilsson L. A. (1994) How pollinator-mediated mat-
ing varies with population size in plants. Oecologia 100: 451–
grant from JSPS Research for the future Program
462.
(00L0162), Japan Ministry of Education, and Science and Goka K. & Takafuji A. (1992) Enzyme variations among Japanese
Culture Grant-in-Aid for Scientific Research (B2– populations of the two-spotted spider mite, Tetranychus urti-
11440227). cae Koch. Applied Entomology and Zoology 27: 141–150.
Goka K. & Takafuji A. (1995) Genetic basis of esterase, malate
References dehydrogenase and phosphoglucoisomerase allozymes in the
two-spotted spider mite, Tetranychus urticae Koch. Applied
Ågren J. (1996) Population size, pollinator limitation, and seed Entomology and Zoology 30: 529–535.
set in the self-compatible herb Lythrum salicaria. Ecology 77: Goodell K., Elam D. R., Nason J. D. & Ellustrand N. C. (1997)
1779–1790. Genetic flow among small populations of a self-compatible
Baker H. G. (1966) The evolution, functioning and breakdown of plant: an interaction between demography and genetics.
heteromorphic incompatibility systems. I. The Plumbagi- American Journal of Botany 84: 1362–1371.
naceae. Evolution 20: 349–368. Handel S. N. (1983) Pollination ecology, plant population struc-
Barrett S. C. H. & Kohn J. R. (1991) Genetic and evolutionary ture, and gene flow. In: Real L. (ed.). Pollination Biology. Aca-
consequences of small population size in plants: implications demic Press, New York, pp. 163–211.
for conservation. In: Falk D. A. & Holsinger K. E. (eds). Genet- Heywood J. S. (1991) Spatial analysis of genetic variation in plant
ics and Conservation of Rare Plants, Oxford University Press, populations. Annual Review of Ecology and Systematics 22: 335–
Oxford, pp. 3–30. 355.

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
120 A . WATA N A B E E T A L .

Heywood J. S. (1993) Biparental inbreeding depression in the Montalvo A. M. (1992) Relative success of self and outcross pol-
self-incompatible annual plant Gaillardia pulchella (Asterceae). len comparing mixed- and single-donor pollinations in Aqui-
American Journal of Botany 80: 545–550. legia caerulea. Evolution 46: 1181–1198.
Husband B. C. & Schemske D. W. (1996) Evolution of the mag- Morán-Palma P. & Snow A. A. (1997) The effect of interplant
nitude and timing of inbreeding depression in plants. Evolu- distance on mating success in federally threatened, self-
tion 50: 54–70. incompatible Hymenoxys herbacea = H. acaulis var. glabra
Johnston M. O. (1993) Tests of two hypothesis concerning pollen (Asteraceae). American Journal of Botany 84: 233–238.
competition in a self-compatible, long-styled species (Lobelia Nishihiro J. & Washitani I. (1998) Effects of population spatial
cardinalis: Lobeliaceae). American Journal of Botany 80: 1400– structure on pollination and seed set of a clonal distylous
1406. plant, Persicaria japonica (Polygonaceae). Journal of Plant
Klinkhamer P. G. L. & de Jong T. J. (1990) Effects of plant size, Research 111: 547–555.
plant density and sex differential nectar reward on pollinator Nishihiro J., Washitani I., Thomson J. D. & Thomson B. A. (2000)
visitation in the protandrous Echium valgare (Boraginaceae). Patterns and consequences of stigma height variation in a
Oikos 57: 399–435. natural population of a distylous plant, Primula sieboldii.
Kohn J. R. (1988) Why be female? Nature 335: 431–433. Functional Ecology 14: 502–512.
Kunin W. E. (1997) Population biology and rarity: on the com- Pasonen H.-L., Pulkkinen P. & Käplylä M. (2001) Do pollen
plexity of dependence in insect - plant interactions. In: Kunin donors with fastest-growing pollen tubes sire the best off-
W. E. & Gaston K. J. (eds). The Biology of Rarity. Chapman & spring in an anemophilous tree, Betula pendula (Betulaceae)?
Hall, London, pp. 150–173. American Journal of Botany 88: 854–860.
Lamont B. B., Klinkhamer P. G. L. & Witkowski E. T. F. (1993) Schoen D. J. (1983) Relative fitness of selfed and outcrossed prog-
Population fragmentation may reduce fertility to zero in eny in Gilia achilleifolia (Polemoniaceae). Evolution 37: 292–301.
Banksia goodii – a demonstration of the Allee effect. Oecologia Sih A. & Baltus M.-S. (1987) Patch size, pollinator behavior, and
94: 446–450. pollinator limitation in Catnip. Ecology 68: 1679–1690.
Levin D. A. (1984) Inbreeding depression and proximity- Silvertown J. W. & Lovett Doust J. (1993) Introduction to Plant
dependent crossing success in Phlox drummondii. Evolution 38: Population Biology. Blackwell Scientific Publications, Oxford.
116–127. Sobrevila C. (1988) Effects of distance between pollen donor and
Levin D. A. (1989) Proximity-dependent cross-compatibility in pollen recipient on fitness components in Espeletia schultzii.
Phlox. Evolution 43: 1114–1116. American Journal of Botany 75: 701–724.
Li C. C. & Horvitz D. G. (1953) Some methods of estimating the Soltis D. E., Haufler C. H., Darrow D. C. & Gastony G. J. (1983)
inbreeding coefficient. American Journal of Human Genetics 5: Starch gel electrophoresis of ferns: acomplication of griding
107–117. buffers, gel and electrode buffers, and staining schedules.
Marshall D. L. (1990) Non-random mating in a wild radish, American Fern Journal 73: 9–27.
Raphanus sativus. Plant Species Biology 5: 143–156. Sowig P. (1989) Effects of plant’s patch size on species composi-
Marshall D. L. (1991) Nonrandom mating in wild radish: varia- tion of pollinator communities, foraging strategies, and
tion in pollen donor success and effects of multiple paternity resource partitioning in bumblebees (Hymenoptera: Adidae).
among one- to six-donor pollinations. American Journal of Bot- Oecologia 78: 550–558.
any 78: 1401–1418. Stephanson A. G. (1981) Flower and fruit abortion: proximate
Marshall D. L., Avritt J. J., Shaner M. & Saunders R. L. (2000) causes and ultimate functions. Annual Review of Ecology and
Effects of pollen load size and composition on pollen donor Systematics 12: 253–279.
performance in wild radish, Raphanus sativus (Brassicaceae). Tanksley S. D. (1979) Linkage, chromosomal association, and
American Journal of Botany 87: 1619–1627. expression of Adh-1 and Pgm-2 in tomato. Biochemical Genet-
Marshall D. L. & Diggle P. K. (2001) Mechanisms of differential ics 3: 35–54.
pollen donor performance in wild radish, Raphanus sativus Thomson J. D. Using pollination deficits to infer pollinator
(Brassicaceae). American Journal of Botany 88: 242–257. declines: can theory guide us? Conservation Ecology [online]
Marshall D. L. & Ellstrand N. C. (1986) Sexual selection in Rapha- 2001; 5 (1): 6. [Cited 13 May 2002] Available from URL:
nus sativus: experimental data on nonrandom fertilization, http://www.consecol.org/vol5/iss1/art6
maternal choice, and consequences of multiple paternity. Turner M. E., Stephens J. C.& Anderson W. W. (1982) Homozy-
American Naturalist 127: 446–461. gosity and patch structure in plant population as a result of
Matsumura C. & Washitani I. (2000) Effects of population size nearest-neighbor pollination. Proceedings of National Academy
and pollinator limitation on seed-set of Primula sieboldii pop- of Science, USA 79: 203–207.
ulations in a fragmented landscape. Ecological Research 15: Waser N. M. & Price M. V. (1983) Optimal and actual outcrossing
307–322. in plants, and the nature of plant–pollinator interactions.
Matsumura C. & Washitani I. (2002) Heterostylous morph differ- In: Jones, C. E. & Little, R. J. (eds). Handbook of Experimental
ences in pollen transfer and deposition patterns in Primula Pollination Biology. Van Nostrand Reinhold, New York, pp.
sieboldii on a visitation by a queen bumblebee, measured with 341–359.
a semi-natural experimental system. Plant Species Biology 17: Washitani I. (1996) Predicted genetic consequences of strong fer-
1–12. tility selection due to pollination loss in an isolated popula-
Menges E. S. (1991) Seed germination percentage increases with tion of Primula sieboldii. Conservation Biology 10: 59–64.
population size in a fragmented prairie species. Conservation Washitani I. & Kabaya H. (1988) Germination responses to tem-
Biology 5: 158–164. perature responsible for the seedling emergence seasonality

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121
 P O P U L AT I O N S PAT I A L E F F E C T S O N S E E D S E T 121

of Primula sieboldii E. Morren in its natural habitat. Ecological Washitani I., Okayama Y., Sato K., Takahashi H. & Ogushi T.
Research 3: 9–20. (1996) Spatial variation in female fertility related to interac-
Washitani I., Kato M., Nishihiro J. & Suzuki K. (1994b) Importance tions with flower consumers and pathogens in a forest metap-
of queen bumble bees as pollinators facilitating inter-morph opulation of Primula sieboldii. Researches on Population Ecology
crossing in Primula sieboldii. Plant Species Biology 9: 169–176. 38: 249–256.
Washitani I., Namai H., Osawa R. & Niwa M. (1991) Species Washitani I., Osawa R., Namai H. & Niwa M. (1994a) Patterns of
biology of Primula sieboldii for the conservation of its lowland- female fertility in heterostylous Primula sieboldii under severe
habitat population: 1. Inter-clonal variations in the flowering pollinator limitation. Journal of Ecology 82: 571–579.
phenology, pollen load and female fertility components. Plant Wright S. (1922) Coefficients of inbreeding and relationship.
Species Biology 6: 27–37. American Naturalist 56: 330–338.

© 2003 The Society for the Study of Species Biology Plant Species Biology 18, 107–121