PLANT-ANIMAL INTERACTIONS - ORIGINAL RESEARCH

How do pollinator visitation rate and seed set relate to species’floral traits and community context?

Amparo Lazaro • Anna Jakobsson •

Ørjan Totland

Received: 15 June 2012 / Accepted: 29 March 2013

� Springer-Verlag Berlin Heidelberg 2013

Abstract Differences among plant species in visitation

rate and seed set within a community may be explained

both by the species’ floral traits and the community con-

text. Additionally, the importance of species’ floral traits

vs. community context on visitation rate and seed set may

vary among communities. In communities where the pol-

linator-to-flower ratio is low, floral traits may be more

important than community context, as pollinators may have

the opportunity to be choosier when visiting plant species.

In this study we investigated whether species’ floral traits

(flower shape, size and number, and flowering duration)

and community context (conspecific and heterospecific

flower density, and pollinator abundance) could explain

among-species variation in visitation rate and seed set. For

this, we used data on 47 plant species from two Norwegian

plant communities differing in pollinator-to-flower ratio.

Differences among species in visitation rate and seed set

within a community could be explained by similar vari-

ables as those explaining visitation rate and seed set within

species. As expected, we found floral traits to be more

important than community context in the community with a

lower pollinator-to-flower ratio; whereas in the community

with a higher pollinator-to-flower ratio, community context

played a bigger role. Our study gives significant insights

into the relative importance of floral traits on species’

visitation rate and seed set, and contributes to our under-

standing of the role of the community context on the fitness

of plant species.

Keywords Conspecific flower density � Display traits �Heterospecific flower density � Inter-specific comparison �Pollinator abundance

Introduction

Plant species differ substantially in the number of polli-

nator visits they receive and in their seed set. Floral traits

are supposed to favour pollination both by attracting

pollinators and by maximizing pollinator efficiency (Fae-

gri and van der Pijl 1966). For this reason, differences

among species in floral traits may be partially responsible

for among-species differences in attractiveness and fitness.

In addition, the community context experienced by indi-

viduals, i.e. the abundance of pollinators and co-flowering

con- and heterospecific plants, is known to affect polli-

nation success (Rathcke 1983) and therefore, it may also

influence among-species differences in visitation rate and

seed set.

The floral traits of a species can be dissected into

components of floral design (e.g. flower shape, size and

number) and temporal characteristics (e.g. flowering

Communicated by Katherine Gross.

Electronic supplementary material The online version of thisarticle (doi:10.1007/s00442-013-2652-5) contains supplementarymaterial, which is available to authorized users.

A. Lazaro (&)

Department of Biodiversity and Conservation, Mediterranean

Institute for Advanced Studies (UIB-CSIC), C/Miquel

Marques 21, Esporles, Balearic Islands 07190, Spain

e-mail: amparo.lazaro@imedea.uib-csic.es

A. Jakobsson

Department of Plant Ecology and Evolution, Uppsala University,

Norbyvagen 18d, 752 36 Uppsala, Sweden

Ø. Totland

Department of Ecology and Natural Resource Management,

Norwegian University of Life Sciences, P.O. Box 5003,

1432 As, Norway

123

Oecologia

DOI 10.1007/s00442-013-2652-5

duration). Intra-specific studies have shown that all of

these floral traits can influence visitation rate and fruit/

seed set. Flower shape can affect the frequency of visitors

(Gomez et al. 2009), and flower size is often positively

related to visitation rate (Bell 1985; Galen and Newport

1987; Eckhart 1991; Dudash et al. 2011) and seed pro-

duction (Galen and Newport 1987). Moreover, the total

number of visits to individuals usually increases with their

flower number, but in a decelerating manner (Robertson

and Macnair 1995; Ohashi and Yahara 2002; Mitchell

et al. 2004). Because of this, the visitation rate and seed

set per flower can be constant (Robertson and Macnair

1995; Goulson et al. 1998; Mitchell et al. 2004) or even

decrease with flower number (Klinkhamer and de Jong

1990; Grindeland et al. 2005). A negative relationship

between flowering duration and visitation rate is plausi-

ble, because long flower longevity and a long flowering

period are often related to a shortage of pollinator visits

and stochastic climatic conditions, respectively (Arroyo

et al. 1981; Blionis et al. 2001). Similar relationships

between floral traits and visitation rate and seed set to

those found within species could also be expected to

occur at the among-species level. However, if these

relationships exist they may be weaker, since plant spe-

cies also differ in other relevant characteristics that

influence visitation rate and seed set, such as reward

levels and the capability of self-fertilization. Lastly,

because morphologically specialized flowers tend to be

also ecologically specialized (Fenster et al. 2004), dif-

ferences among species in flower shape may influence

visitation frequencies (McCall and Primack 1992) and can

also modify the relationships between the variables

described above and visitation rate and seed set. To our

knowledge, only a couple of studies at the community

Table 1 The plant species (Mossberg and Stenberg 2007) of each community included in this study, their family, the observation time (Obs.time; hours) and the number of individuals sampled for seed set (Sam. ind.)

Community Species Family Obs.

time

Sam.

ind.

Community Species Family Obs.

time

Sam.

ind.

Ryghsetra Anthyllis vulneraria Fabaceae 6.3 26 Finse Astragalus alpinus Fabaceae 21.3 26

Centaurea jacea Asteraceae 11.3 36 Bartsia alpina Scrophulariaceae 20.7 44

Centaurea scabiosa Asteraceae 3.0 31 Campanula rotundifolia Campanulaceae 3.7 36

Fragaria vesca Rosaceae 9.0 28 Cerastium alpinum Caryophyllaceae 11.3 33

Galium boreale Rubiaceae 5.0 23 Dryas octopetala Rosaceae 9.0 34

Galium mollugo Rubiaceae 20.0 55 Erigeron uniflorus Asteraceae 15.3 50

Geranium sylvaticum Geraniaceae 7.0 15 Euphrasia frigida Scrophulariaceae 6.0 86

Geum rivale Rosaceae 5.7 15 Gentiana nivalis Gentianaceae 4.3 41

Hieracium sec. hieracium Asteraceae 1.0 15 Gentianella campestris var. suecica Gentianaceae 1.3 23

Knautia arvensis Dipsacaceae 3.0 15 Geranium sylvaticum Geraniaceae 10.3 23

Lathyrus linifolius Fabaceae 3.0 15 Leontodon autumnalis Asteraceae 22.3 34

Leucanthemum vulgare Asteraceae 17.0 66 Parnassia palustris Celastraceae 19.7 80

Linum catharticum Linaceae 9.0 71 Pinguicula vulgaris Lentibulariaceae 4.0 20

Lotus corniculatus Fabaceae 8.0 17 Potentilla crantzii Rosaceae 15.0 40

Pilosella lactucella Asteraceae 6.3 34 Pyrola minor Ericaceae 3.0 16

Pilosella officinarum Asteraceae 10.3 27 Ranunculus acris Ranunculaceae 18.7 29

Pilosella pubescens Asteraceae 6.3 41 Saussurea alpina Asteraceae 5.3 44

Polygala vulgaris Polygalaceae 12.0 48 Silene acaulis Caryophyllaceae 20.3 49

Potentilla erecta Rosaceae 24.7 51 Tofieldia pusilla Liliaceae 8.0 32

Potentilla thuringiaca Rosaceae 7.3 24 Veronica alpina Plantaginaceae 7.0 53

Primula veris Primulaceae 5.7 51 Viscaria alpina Caryophyllaceae 0.7 15

Prunella vulgaris Lamiaceae 2.3 15

Ranunculus acris Ranunculaceae 7.7 15

Trifolium medium Fabaceae 5.3 16

Trifolium pratense Fabaceae 10.7 17

Vicia cracca Fabaceae 4.3 15

Oecologia

123

level have related species’ floral traits to visitation rates

(McCall and Primack 1992; Hegland and Totland 2005);

and these studies have not examined the relationship

between species’ floral traits and seed set.

In addition to species’ floral traits, the community

context, i.e. the plants and pollinators occurring simulta-

neously in the community, may also influence visitation

rate and seed set through competitive or facilitative inter-

actions (Rathcke 1983). The local plant density of a species

is often positively related to its visitation rate (Klinkhamer

and de Jong 1990; Feinsinger et al. 1991; Kunin 1993;

Makino et al. 2007) and seed set (Feinsinger et al. 1991;

Jennersten and Nilsson 1993; Kunin 1993), because poll-

inators preferentially visit high-density patches (Stephens

and Krebs 1986; Goulson 2003; Hersch and Roy 2007)

and/or because there is an increased likelihood of receiving

outcross and compatible pollen at high conspecific density

(Jennersten and Nilsson 1993; Karron et al. 1995). The

abundance of other blooming species can have either

positive or negative effects on pollinator visitation rate and

seed set of a species (Schemske 1981; Rathcke 1983;

Feinsinger et al. 1991; Moeller 2004). Moreover, the

abundance of adequate pollinators during the flowering

period of a species may be positively related to its visita-

tion rate and seed set (Kudo 1993; Kwak and Bergman

1996; Price et al. 2005).

The distinction between the effect of species’ floral

traits and the community context is relevant in ecology

and evolution, because while species’ floral traits are

supposed to have evolved to favour pollination, the con-

text in which they exist is fortuitous. However, to date

little is known about the relative importance of commu-

nity context and species’ floral traits on visitation rate and

seed set. Moreover, the effects of floral traits and local

community context on visitation rate often vary among

plant populations (Eckhart 1991; Lazaro and Totland

2010) and insect taxa (Thompson 2001; Lazaro et al.

2009). Therefore, it is conceivable that the relative

importance of species’ floral traits versus community

context on visitation rate and seed set also vary among

communities, depending on the abundance of their poll-

inators. In communities where the pollinator-to-flower

ratio is low, competition theory predicts a low competi-

tion among pollinators for floral resources (Alley 1982);

and pollinators may be choosier when deciding which

plant species to visit (Schmitt 1983). In such communities

we expect floral traits to be more important than com-

munity context. Conversely, in communities with high

pollinator-to-flower ratio (i.e. where pollinators may

compete for floral resources), pollinator competence may

alter initial flower preferences (e.g. Chittka and Thomson

2001), and pollinators would not have the opportunity to

be choosy (Schmitt 1983). As a result, pollinator choices

of plant species may be more influenced by the commu-

nity context.

In this study, we investigated the relative importance of

species’ floral traits and community context on visitation

rate and seed set, using data from two communities in

southern Norway: one semi-natural species-rich meadow

with a high abundance and diversity of pollinators, and one

alpine community where pollinators are at a lower abun-

dance. Our specific objectives were: (1) to investigate how

pollinator visitation rate and seed set relate to species’

floral traits (flower shape, size and number, and flowering

duration) and community context (conspecific and hetero-

specific flower density, and pollinator abundance) in the

two study communities; and (2) to determine whether the

importance of species’ floral traits vs. community context

differed between the two study communities; we expected

less importance of the community context in the commu-

nity where pollinators are scarce.

Materials and methods

Study areas and species

We conducted our study in two plant communities of

southern Norway. The first community is a species-rich

semi-natural meadow at Ryghsetra (59�4400300N,

10�0204800E), Buskerud county, at 300 m altitude. Here, the

blooming season starts in early-mid May and terminates in

mid-late August, and around 55 species are in flower dur-

ing this period. The flower visitor assemblage of this

community consisted of 72.7 % bumblebees, 11 % mus-

coid flies, 5.5 % solitary bees, 4.7 % hover flies, 2.4 %

ants, 1.6 % butterflies, 0.9 % honeybees, 0.5 % beetles,

0.5 % beeflies, and 0.04 % wasps in the study year. The

average (across all the species) number of pollinator visits

recorded per hour was 64.7, and the pollinator visits to

flowers ratio was 0.17.

The second community is a southwest exposed slope on

Sandalsnuten, at Finse (*60�N, 7�E), in southwest alpine

Norway, at 1,450 m altitude. Here, the blooming season

starts in late June and terminates in late August, and ca. 25

species are in flower during this period. The flower visitor

assemblage of this community consisted of 85.8 % mus-

coid flies, 7.9 % butterflies, 3.6 % bumblebees and 2.5 %

hover flies in the study year. The average number of pol-

linator visits recorded per hour was 47.9, and the pollinator

visits to flowers ratio was 0.09. Thus, at Finse there is

substantially more flowers relative to flower visits, com-

pared to Ryghsetra.

We included in this study the 47 plant species (26 from

Ryghsetra and 21 from Finse) for whichwe obtained ade-

quate sample sizes to conduct the analyses (Table 1).

Oecologia

123

Pollinator observations and flower densities

We observed insect visitation to plant species from 13 May

to 19 July 2006 at Ryghsetra, and from 3 July to 5 August

2006 at Finse. These periods cover the whole flowering

period of most plant species in these communities. At the

beginning of the field season, we placed 30 permanent

2 9 2-m plots in both communities. The inner 1 9 1-m

square of these plots was marked and observations were

conducted within it. Due to the particular characteristics of

the study sites, and in order to get the best representation of

each community, we placed the plots at random at Finse,

and systematically at Ryghsetra. Twenty additional non-

permanent plots were located at Ryghsetra during June

(flowering peak of the community). The location of these

additional plots was randomly changed after each obser-

vation period. This was necessary in order to collect

enough information on the visits to all the species bloom-

ing during the flowering peak of the community. These

differences in sampling strategy likely had no influence on

our results, since we focus on among-species comparisons

within communities.

Observations were conducted every day with weather

conditions allowing pollinator activity, between 0900 and

1800 hours, using 20-min observation periods. The order

of observation of plots was random, but we never

observed the same plot more than once per day. During

each observation period we noted the number of flower

visitors to flowers or inflorescences (depending on the

species) of all flowering species in the plots. A visit was

defined to have occurred when the visitor’s body made

contact with the reproductive organs (stigma or anther) of

the flower. Flower visitors that contacted reproductive

organs of plants were regarded as pollinators, although

some of these visitors may be ineffective pollinators. In

total, we conducted 118 and 134 observation periods at

Ryghsetra and Finse, respectively (see Table 1 for total

number of observation hours for each study plant

species).

After each observation period, we counted the number

of open flower units of each species occurring both in the

2 9 2-m plots and in the 1 9 1-m inner squares. The

number of flower units counted in the 1 9 1-m inner

squares was used to calculate flower visitation rates,

whereas the number of flower units counted in the 2 9 2-m

plots was used to calculate floral abundances (see below).

The flower units were normally flowers, except in the case

of species with pseudanthia (e.g. Asteraceae, Knautia

arvensis), for which the flower unit was the inflorescence

[Hegland and Totland 2005; see Table S1, Electronic

Supplementary Material (ESM) for flower units used for

each species].

Visitation rate

For each species, we calculated a visitation rate per flower

unit in each 20-min observation period by dividing the

number of observed pollinator visits by the number of

observed open flower units. The values per observation

period were averaged to obtain species’ means (hereafter

‘visitation rate’).

Seed set

We haphazardly selected and marked from one to three

individuals per plant species in each plot where they

occurred, counted their total number of flowers and

inflorescences, and marked one flower unit (Table S1,

ESM) on each individual. For plant species with low

abundance (B30 individuals) in the permanent plots, we

haphazardly selected 15 additional individuals outside the

plots to increase sample size for the study of seed set.

When fruits were ripe we counted the total number of fruit

units and collected the fruit units of the marked flower

units. Fruit units were fruits or infructescences depending

on the species (see Table S1, ESM, for fruit units for each

species). The number of undeveloped seeds, aborted seeds

and fully developed seeds of these fruit units were counted

in the lab. We calculated fruit set as the proportion of

female flower units that developed fruit units. In order to

use a standardized measure of fecundity that was compa-

rable among plant species and that included both the

production of fruits and seed set per fruit, we estimated

total seed set at the plant level. This total seed set (here-

after ‘seed set’) was calculated by dividing the total

number of developed seeds produced by the plant (fruit

set 9 average number of developed seeds per fruit

unit 9 flower units) by the total number of ovules pro-

duced by the plant (average number of ovules per flower

unit 9 number of flower units). We were unable to count

the number of undeveloped seeds in Primula veris, so in

order to estimate the total number of ovules produced by a

plant, we multiplied its number of flowers by 50, which is

the average number of ovules in this species (Wissman

2006). The values per individual plant were averaged to

obtain species’ means.

Species’ floral traits

All the floral traits of the species included in this study were

calculated at the flower level, except for the species with

pseudanthia (i.e. inflorescences with a flower-like struc-

ture), for which the pseudanthium was used instead. For

each plant species we used the same flower units as those

used for the estimation of visitation rates and abundances

Oecologia

123

(Table S1, ESM). For the description of floral traits we used

the term ‘flower’ to refer to flowers or pseudanthia.

Flower shape

To describe flower shape in a simple manner that allows

statistical analyses, we categorised species into ‘unspe-

cialized’ and ‘specialized’ flowers, following the termi-

nology by Larson and Barrett (2000). Unspecialized

flowers could potentially be visited by a wide range of

pollinator groups, and included open flowers (e.g. Poten-

tilla erecta) and small clusters of flowers in open inflo-

rescences (e.g. Asteraceae). Specialized flowers were those

that exerted a difficulty of handling, and consequently

could only be accessed by a subset of the pollinator fauna.

Among the specialized flowers we included closed flowers

(e.g. Fabaceae), semi-closed flowers (e.g. Scrophularia-

ceae) and pendular flowers (e.g. Geum rivale). Table S1

(ESM) shows the categories of flower shape used for the

study species.

Flower size

We measured tube length and width of flowers on all

species except those with pseudanthia (e.g. Asteraceae,

Knautia arvensis), for which we measured width and

length of inflorescences. Measurements (to the nearest

millimetre) were conducted in the field on 30 randomly

selected individuals per species (one flower/pseudanthia

per individual), using a digital calliper. In order to obtain a

value for flower size that integrated all these measure-

ments, we calculated the area of the flower, following

Hegland and Totland (2005). Thus, for visual displays with

circular outline (e.g. Leucanthemum vulgare) we used the

formula pr2, where r = radius; and for visual displays with

a length dimension (e.g. Lathyrus linifolius) we used the

formula of a cylinder (without a base) 2prl ? pr2, where

l = length. We used these areas for visual displays of

different shape to take into account how insects approach

the flowers (either from the upper part of them, or from

different heights and angles). Table S1 (ESM) shows the

flower size calculated for each study species.

Flower number

We calculated flower number for each species as the

average number of flowers per individual.

Flowering duration

We estimated flowering duration as the number of days

each species flowered, from the first to the last day we

observed open flowers of the species.

Community context

Conspecific flower density

We estimated conspecific flower density as the average

number of flower units of each species per square metre

(only the 2 9 2-m permanent plots were included in this

calculation). Although conspecific flower density is a

population variable, we included it as a community context

variable, since with our variable classification, we intended

to discriminate between the traits of individuals within

species and the variables related to the context of

individuals.

Heterospecific flower density

We calculated heterospecific flower density in each com-

munity during the flowering period of each study species.

For this, we averaged the number of open heterospecific

flower units (i.e. all except the focal species) recorded per

plot in the whole community while the focal species

flowered.

Pollinator abundance

We calculated pollinator abundance for each study species

by using the average number of pollinator visits recorded

per observation period in the whole community during the

flowering period of the study species.

Statistical analyses

To study the relative importance of species’ floral traits and

community context for pollinator visitation rate and seed

set, we used generalized linear models separately for each

community and response variable. In these analyses the

plant species were the sample units and visitation rate and

seed set the response variables. Due to the nature of the

data, we used: (1) gamma distributions and inverse link

functions for the analyses of visitation rate (visitation rate

?1; in order to avoid zeros in the response variable); and

(2) binomial distributions and logit link functions for the

analyses of seed set. We used all the variables described in

Species’ floral traits and Community context as predictors.

We included visitation rate as an additional predictor in the

models of seed set to account for the potential direct effect

of visitation rate on seed set (Jennersten 1988; Pauw 2007),

that is independent on its effect through floral traits or

community context. We also included sampling effort (i.e.

the number of hours we observed plots containing each

species) as an additional predictor variable in the analyses

of visitation rate. We conducted the analyses of visitation

rate with and without the plant species observed \3 h

Oecologia

123

(Table 1), but since the results did not change (results not

shown), we present here the results using all the study

species. Prior to analyses, we ran variation inflation factor

(VIF) analyses to identify collinear predictor variables that

should be removed from further analyses (Zuur et al.

2009). VIF values were smaller than 3 for all variables, so

none of the predictors needed to be removed (Zuur et al.

2009). We used automatic selection of best models

(package glmulti, R 12.2) using corrected Akaike infor-

mation criterion to select the best models (Calcagno and de

Mazancourt 2010). All analyses were conducted in R 12.2

(R Development Core Team 2008). Means are accompa-

nied by their SE throughout the text.

Results

Visitation rate

At Ryghsetra, several variables describing species’ floral

traits significantly explained variation among species in

visitation rate (Table 2). Visitation rate was higher for

species with unspecialized flowers compared to species

with specialized flowers (1.34 ± 0.15 vs. 0.13 ± 0.01

visits/20 min, respectively), and decreased with flower

number (Table 2; Fig. 1a). Visitation rate was also

related to flowering duration, but this relationship dif-

fered between species with unspecialized and specialized

flowers (Table 2): while visitation rate was negatively

related to flowering duration in species with unspecial-

ized flowers, there was no relationship between flowering

duration and visitation rate for species with specialized

flowers (Fig. 1b). In addition, visitation rate was signif-

icantly related to two variables describing the community

context: pollinator abundance and heterospecific flower

density (Table 2). Visitation rate increased with pollina-

tor abundance in the community consistently for both

specialized and unspecialized flowers (Table 2; Fig. 1c);

however, the relationship between visitation rate and

heterospecific flower density depended on flower shape.

Visitation rate decreased with heterospecific flower

density in species with unspecialized flowers, whereas

visitation rate of species with specialized flowers was

not related to heterospecific flower density (Table 2;

Fig. 1d).

At Finse, visitation rate was also higher for species with

unspecialized flowers compared to species with specialized

flowers (0.57 ± 0.04 vs. 0.15 ± 0.03 visits/20 min,

respectively; Table 2), and increased with flower size

(Table 2; Fig. 2). None of the variables describing the

community context significantly explained variation

among species in visitation rate in this community

(P [[ 0.05).

Seed set

At Ryghsetra, several variables describing species’ floral

traits significantly explained variation among species in

their seed set. Seed set was higher for species with

unspecialized flowers compared to species with specialized

flowers (0.46 ± 0.01 vs. 0.25 ± 0.02 seeds/ovules,

respectively; Table 3) and decreased with flower number

(Table 3; Fig. 3a). Flowering duration was also related to

seed set; however, the direction of the relationship between

flowering duration and seed set was different for species

with specialized and unspecialized flowers (Table 3;

Fig. 3b). While there was a positive relationship between

flowering duration and seed set in species with unspecial-

ized flowers, there was a negative relationship between

these two variables in species with specialized flowers

(Fig. 3b). In addition, variables describing the community

context also significantly explained variation among spe-

cies in their seed set. Thus, seed set increased with polli-

nator abundance (Fig. 3c), conspecific flower density

(Fig. 3d) and heterospecific flower density (Fig. 3e). The

relationship between seed set and heterospecific flower

density was, however, stronger in species with specialized

than unspecialized flowers (Table 3; Fig. 3e).

At Finse, seed set was higher in species with specialized

than unspecialized flowers (0.64 ± 0.04 vs. 0.58 ± 0.03

seeds/ovules, respectively; Table 3). Seed set decreased

with flower number (Fig. 4a), and flowering duration

(Fig. 4b), whereas it increased with flower size (Fig. 4c).

None of the variables describing the community context

significantly explained variation among species in seed set

at Finse (P � 0.05).

Discussion

We have shown here that differences among species in

visitation rate and seed set within a community can be

explained by similar variables to those explaining visitation

rate and seed set within species. In our study, species’ floral

traits (flower shape, size and number, and flowering dura-

tion), significantly explained variation in visitation rates and

seed set among species in both communities. However, as

expected, the variables describing the community context

(conspecific and heterospecific flower density and pollina-

tor abundance) were significantly related to visitation rate

and seed set only at Ryghsetra, the community where the

pollinator-to-flower ratio was higher. Although the vari-

ables that explained variation in visitation rate also

explained variation in seed set, the best models for seed set

were more complex and included more variables. This

indicates that the relationship between visitation and seed

production is not straightforward (as shown also by the lack

Oecologia

123

of significance of the variable visitation rate in the models

of seed set). Such decoupling between visitation rate and

seed set is not surprising, and can occur if species can

compensate for pollen limitation through self-fertilization

(Karoly 1992; Kalisz et al. 2004), if they differ in the

number of pollinator visits needed to fertilize the ovules

(Molano-Flores et al. 1999), or when resource acquisition

limits seed production or maturation (e.g. Totland 2001).

Moreover, an absence of a direct link between visitation and

seed production can be due to inter-specific pollen transfer

(Galen and Gregory 1989; Caruso and Alfaro 2000; Ja-

kobsson et al. 2007), and/or pollinator behaviour that dif-

ferentially affect the quantity and quality of flower visits

(e.g. de Jong et al. 1993; Klinkhamer and de Jong 1993).

Table 2 Best models explaining inter-specific variation in visitation rate at Ryghsetra and Finse, using gamma distributions and inverse link

functions, and corrected Akaike information criterion (AICc) for model selection

Community Variable Estimate SE t P-value

Ryghsetra Intercept 1.04 0.17 5.95 \0.0001

Shape (unspecialized) -0.87 0.20 -4.29 0.0004

Flower number 0.004 0.0008 4.67 0.0002

Pollinator abundance -0.008 0.001 -5.65 \0.0001

Flowering duration -0.003 0.004 -0.57 0.57

Heterospecific flower density -0.003 0.005 -0.53 0.61

Shape (unspecialized) 9 flowering duration 0.01 0.005 2.08 0.052

Shape (unspecialized) 9 heterospecific flower density 0.01 0.006 2.54 0.021

Finse Intercept 1.13 0.13 8.96 \0.0001

Shape (unspecialized) -0.30 0.11 -2.85 0.01

Flower size -0.001 0.0003 -3.53 0.002

Note that the sign of the parameter estimates is opposite to the direction of the relationship between variables due to the use of an inverse link

function

Fig. 1 Relationships between

visitation rate at Ryghsetra and

a flower number, b flowering

duration, c pollinator abundance

and d heterospecific flower

density. Lines represent the

estimates of the best model and

the circles represent the partial

residuals of the model. When

there is a significant interaction,

the relationship is shown for

species with unspecialized

(empty circles, dotted lines) and

specialized flowers (filledcircles, solid lines)

Oecologia

123

Species’ floral traits

In both communities, species with unspecialized flowers

had higher visitation rates than those with specialized

flowers; this is in line with the findings of McCall and

Primack (1992) who showed higher visitation rates to open

flowers compared to tubular flowers. As expected, flower

shape also modified the relationships between other pre-

dictor variables and visitation rate and seed set.

At Finse, we found that both visitation rate and seed set

increased with flower size. Several studies at the within-

species level have found flower size to positively affect

visitation (Bell 1985; Galen and Newport 1987; Eckhart

1991; Dudash et al. 2011) and seed production (Galen and

Newport 1987), and this effect has mainly been attributed

to a higher attractiveness of larger flowers. However,

higher seed production in larger flowers could also reflect

resource acquisition rather than increased pollination suc-

cess (Harder and Johnson 2009). To our knowledge, our

study is the first to report relationships between flower size

and visitation rate and seed set at the among-species level.

However, we only found these relationships in the alpine

community, Finse. The greater importance of flower size in

the alpine community can be due to a stronger plant species

selectivity by insects there (Schmitt 1983) and/or a larger

similarity among plant species in other floral traits [most

often unspecialized (Table 2) and with single or few

flowers]. Future studies may shed light on whether the

difference among species in flower size is an important

determinant of visitation rate and seed set in other alpine

communities.

Flower number was negatively related to visitation rate

and seed set in both communities. At the within-species

level, several studies have shown that the total number of

visits to a plant increases with its flower number (Eckhart

1991; Makino et al. 2007), but in a decelerating manner

(Robertson and Macnair 1995; Ohashi and Yahara 2002;

Mitchell et al. 2004). One reason for such a pattern may be

that insects attempt to avoid revisiting previously visited,

and thus less rewarding, flowers (Dreisig 1995). As a

consequence, visitation rate and seed set per flower may be

independent of (Robertson and Macnair 1995; Goulson

et al. 1998; Mitchell et al. 2004), or even decrease with,

flower number (Klinkhamer and de Jong 1990; Grindeland

et al. 2005). This may explain the negative relationship

between flower number and visitation rate and seed set that

we found at the inter-specific level at Ryghsetra. At Finse,

seed set was also negatively related to flower number, but

Fig. 2 Relationship between visitation and flower size at Finse. The

line represents the estimates of the best model and the circlesrepresent the partial residuals of the model

Table 3 Best models explaining inter-specific variation in seed set at Ryghsetra and Finse, using binomial distributions and logit link functions,

and AICc for model selection

Community Variable Estimate SE t P-value

Ryghsetra Intercept -0.71 0.18 -3.90 \0.0001

Shape (unspecialized) -1.28 0.37 -3.45 0.0006

Flower number -0.01 0.005 -20.59 \0.0001

Conspecific flower density 0.18 0.01 15.94 \0.0001

Pollinator abundance 0.03 0.01 5.31 \0.0001

Flowering duration -0.05 0.01 -5.74 \0.0001

Heterospecific flower density 0.05 0.01 7.28 \0.0001

Shape (unspecialized) 9 flowering duration 0.07 0.01 6.94 \0.0001

Shape (unspecialized) 9 heterospecific flower density -0.04 0.01 -4.16 \0.0001

Finse Intercept 1.60 0.07 24.38 \0.0001

Shape (unspecialized) -0.28 0.06 -4.53 \0.0001

Flower number -0.02 0.003 -9.63 \0.0001

Flowering duration -0.06 0.005 -12.93 \0.0001

Flower size 0.01 0.0003 17.91 \0.0001

Oecologia

123

visitation rate was not. Although flower number can be

correlated with seed set due to resource rather than polli-

nation limitation (Harder and Johnson 2009), this possi-

bility seems quite unlikely in this case, because the

relationship found is negative, whereas a higher number of

flowers should indicate more resources in the plant to set

the seeds. Instead, reduced seed set in species with larger

flower numbers at Finse may more likely be due to a

negative effect of increased geitonogamy (Brys and Jac-

quemyn 2010).

The results for flowering duration were complex. At

Ryghsetra, flowering duration was negatively related to

visitation rate, whereas it was positively related to seed set

in species with unspecialized flowers. These results agreed

with our expectations: while total visitation (across the

whole flowering period) and seed set may increase with

flowering duration (Dieringer 1991), visitation rate per

flower might decrease with flowering duration, since spe-

cies with low flower visitation rates may compensate with

long flowering periods or longer flower duration (Arroyo

et al. 1981; Blionis et al. 2001). For specialized species at

Ryghsetra, however, seed set decreased with flowering

duration; and the same occurred with both specialized and

unspecialized species at Finse. These unexpected rela-

tionships could appear if the species that bloom longer do it

partially during ‘inappropriate’ periods. Thus, it could be

Fig. 3 Relationships between

seed set at Ryghsetra and

a flower number, b flowering

duration, c pollinator

abundance, d conspecific flower

density and e heterospecific

flower density. Lines represent

the estimates of the best model

and the circles represent the

partial residuals of the model.

When there is a significant

interaction, the relationship is

shown for species with

unspecialized (empty circles,

dotted lines) and specialized

flowers (filled circles, solidlines)

Oecologia

123

that part of the flowering period of long-flowering spe-

cialized species at Ryghsetra does not overlap with the

phenological timing of their specific pollinators. In the case

of Finse, an early flowering period is known to increase the

probability of completing seed production before the onset

of winter (Totland 1993). Perhaps in this community the

longest flowering species begin setting part of their fruits

too late, when abiotic conditions are not adequate for fruit

and seed maturation. Future studies may confirm whether

this negative relationship is maintained over the years and/

or similar patterns are found in specialized flowers and in

other alpine communities.

Community context

Among-species variation in visitation rate and seed set was

related to the community context variables (conspecific and

heterospecific flower density and pollinator abundance)

only at Ryghsetra, the community with a higher pollinator-

to-flower ratio. We hypothesized that the community con-

text would have a stronger influence on visitation rate and

seed set in communities with a higher pollinator-to-flower

ratio compared to those with a lower pollinator-to-flower-

ratio. Pollinator foraging is determined by energetic

requirements (Stephens and Krebs 1986; Goulson 2003),

and when the pollinator-to-flower ratio is low there may be

low competition among pollinators for flowers (Alley

1982). Pollinators could then be choosier (Schmitt 1983)

and visit those flowers they prefer based on innate prefer-

ences and/or acquired experience (e.g. Chittka and Thom-

son 2001). On the contrary, when the pollinator-to-flower

ratio is high, competition among pollinators may modify the

reward levels and exploitation costs of otherwise preferred

plant species, and pollinators would not have the opportu-

nity to be so choosy (Schmitt 1983). As a result, insect

choices of plant species during foraging might be more

influenced by the scenario imposed by the plant-pollinator

community. Our results, showing a larger influence of

context variables on visitation rate and seed set at Ryghs-

etra, but not Finse, supported our hypothesis. However, our

correlative approach conducted in two communities has

important limitations. First, our failure to detect relation-

ships between the community context and seed set in the

alpine community could also be caused by strong resource

limitation that cancels out these relationships. However, we

find this possibility quite unlikely because significant rela-

tionships between floral traits and seed set are found at

Finse, and because flower density (included among the

community context variables) is one of the factors that can

strongly influence resource acquisition and therefore, seed

set. Second, the lack of replication of sites with high and

low pollinator-to-flower ratios does not allow us to form

strong conclusions about the importance of community

context for plant fitness in communities differing in polli-

nator-to-flower ratio. Although complex studies at the

community level are difficult to replicate in many localities,

the analysis of more communities is desirable to test whe-

ther the results presented here can be generalised.

Fig. 4 Relationships between seed set at Finse and a flower number,

b flowering duration and c flower size. The lines represent the

estimates of the best model and the circles represent the partial

residuals of the model

Oecologia

123

Seed set, but not visitation rate, increased with con-

specific flower density. Other studies have found seed set to

be positively related to conspecific density, through an

increase in the number of visits (Klinkhamer and de Jong

1990; Kunin 1993; Field et al. 2005; Makino et al. 2007)

and/or their quality (i.e. higher number of conspecific and

outcross pollen grains deposited on the stigmas; Jennersten

and Nilsson 1993; Kunin 1993; Bosch and Waser 1999;

Field et al. 2005) with density. Since we did not find any

relationship between conspecific density and visitation rate,

the relationship between conspecific density and seed set

may more likely be due to an increase in the amount of

available conspecific outcross pollen (Jennersten and

Nilsson 1993; Karron et al. 1995) in more abundant species

in this community.

Visitation rate and seed set were also related to the other

two variables describing the community context (hetero-

specific flower density and pollinator abundance) at Ry-

ghsetra. Visitation rate and seed set consistently increased

with pollinator abundance in species with both specialised

and generalised flowers. This was expected because the

higher the abundance of pollinators in the community at the

time a species flowers, the higher the chance a flower of

this species has of being visited and the higher the chance

of this species to set seeds. Lastly, heterospecific flower

density was negatively related to visitation rate in unspe-

cialized flowers, indicating inter-specific competition for

pollinator attraction when the density of co-flowering

species is high (Rathcke 1983). The species with special-

ized flowers were not affected by the heterospecific flower

density, perhaps because they have more loyal specific

pollinators (McCall and Primack 1992; Fenster et al. 2004).

Contrary to this, there was a positive relationship between

heterospecific flower density and seed set for species with

both specialized and unspecialized flowers. It is possible

that even though competition with other co-flowering

species decreased visitation rates, pollinators become

choosier when the resource is abundant (Schmitt 1983;

Lazaro et al. 2009, 2010), segregating their niches and

therefore increasing the quality of each visit (i.e. increasing

the amount of conspecific pollen deposited on the stigmas).

Future work will be needed to explore this idea.

Conclusion

In this study we have shown that the relationships between

floral traits and community context and visitation rate and

seed set, which are commonly found within species, also

exist among species. Moreover, in the community with the

lowest pollinator-to-flower ratio, species’ floral traits had

more influence on visitation rate and seed set than the

community context. We hypothesized a lower effect of

community context on plant fitness when the pollinator-to-

flower ratio is low, because when there is weak competition

between pollinators for flowers, insects might be choosier

as regards the plant species they visit (Schmitt 1983).

However, while our results support this hypothesis, the

inclusion of more communities in the analyses would be

needed to properly test it. The inter-specific relationships

between species’ traits and community context and the

fitness of species are still far from understood and their

study deserves further exploration. Experimental manipu-

lations would help to assess cause-effect relationships that

our correlative data do not allow us to ascertain. Also,

inclusion in the analyses of other species traits such as

capability of self-pollination and flower rewards, and

examination of the interactions that may exist between

predictor variables on visitation rate and seed set (Stanton

et al. 1991; Ohashi and Yahara 2002; Grindeland et al.

2005; Biernaskie and Gegear 2007), would help to improve

future studies on this subject.

Acknowledgments We thank Paul Aakerøy, Amparo Castillo,

Astrid Haavik, Manuel Hidalgo, Kristian Kyed, Rebekka Lundgren,

Kirsten Marthinsen, Anton Perez, Martın Piazzon, Anna Karina

Schmidt, Mari Steinert, Siril Stenerud, Jorum Vallestad, Hartvig

Velund, Silje Wang, Torstein Wilmot, and Magnus Øye for their

invaluable help in the field and the lab. We are also very grateful to

Asier R. Larrinaga and Martin Piazzon for their statistical advice and

to Amanda M. Dorsett for language editing. This study was supported

by the project 170532/V40 financed by the Norwegian Research

Council. During the writing of this manuscript A. L. was supported

by a Juan de la Cierva contract (Spanish Ministry of Economy and

Competitiveness). The sampling complies with the current laws of the

country (Norway) where it was performed.

References

Alley TR (1982) Competition theory, evolution, and the concept of an

ecological niche. Acta Biotheor 31:165–179

Arroyo MTK, Armesto JJ, Villagran C (1981) Plant phenological

patterns in the high Andean cordillera of central Chile. J Ecol

69:205–223

Bell G (1985) On the function of flowers. Proc R Soc Lond B

224:223–265

Biernaskie JM, Gegear RJ (2007) Habitat assessment ability of

bumble-bees implies frequency-dependent selection on floral

rewards and display size. Proc R Soc Lond B 274:2595–2601

Blionis GJ, Halley JM, Vokou D (2001) Flowering phenology of

Campanula on Mt Olympos, Greece. Ecography 24:696–706

Bosch M, Waser NM (1999) Effects of local density on pollination

and reproduction in Delphinium nuttallianum and Aconitumcolumbianum (Ranunculaceae). Am J Bot 86:871–879

Brys R, Jacquemyn H (2010) Floral display size and spatial

distribution of potential mates affect pollen deposition and

female reproductive success in distylous Pulmonaria officinalis(Boraginaceae). Plant Biol 12:597–603

Calcagno V, de Mazancourt C (2010) glmulti: an R package for easy

automated model selection with (generalized) linear models.

J Stat Softw 34:1–29

Caruso CM, Alfaro M (2000) Interspecific pollen transfer as a

mechanism of competition: effect of Castilleja linariaefolia

Oecologia

123

pollen on seed set of Ipomopsis aggregata. Can J Bot 78:

600–606

Chittka L, Thomson JD (2001) Cognitive ecology of pollination:

animal behavior and floral evolution. Cambridge University

Press, Cambridge

De Jong TJ, Waser NM, Klinkhamer PGL (1993) Geitonogamy: the

neglected side of selfing. Trends Ecol Evol 8:321–325

Dieringer G (1991) Variation in individual flowering time and

reproductive success of Agalinis strictifolia (Scrophulariaceae).

Am J Bot 78:497–503

Dreisig H (1995) Ideal free distributions of nectar foraging bumble-

bees. Oikos 72:161–172

Dudash MR, Hassler C, Stevens PM, Fenster CB (2011) Experimental

floral and inflorescence trait manipulations affect pollinator

preference and function in a hummingbird-pollinated plant. Am J

Bot 98:275–282

Eckhart VM (1991) The effects of floral display on pollinator

visitation vary among populations of Phacelia linearis (Hydro-

phyllaceae). Evol Ecol 5:370–384

Faegri K, van der Pijl L (1966) The principles of pollination ecology.

Pergamon Press, Oxford

Feinsinger P, Tiebout HM III, Young BE (1991) Do tropical bird-

pollinated plants exhibit density-dependent interactions? Field

experiments. Ecology 72:1953–1963

Fenster CB, Armbruster WS, Wilson P, Dudash MR, Thomson JD

(2004) Pollination syndromes and floral specialization. Annu

Rev Ecol Evol Syst 35:375–403

Field DL, Ayre DJ, Whelan RJ (2005) The effect of local plant

density on pollinator behaviour and the breeding system of

Persoonia bargoensis (Proteaceae). Int J Plant Sci 166:969–977

Galen C, Gregory T (1989) Interspecific pollen transfer as a

mechanism of competition—consequences of foreign pollen

contamination for seed set in the alpine wild flower Polemoniumviscosum. Oecologia 81:120–123

Galen C, Newport MEA (1987) Bumble bee behaviour and selection

on flower size in the sky pilot, Polemonium viscosum. Oecologia

74:20–23

Gomez JM, Perfectti F, Bosch J, Camacho JPM (2009) A geographic

selection mosaic in a generalized plant-pollinator-herbivore

system. Ecol Monogr 79:245–263

Goulson D (2003) Bumblebees: behaviour and ecology. Oxford

University Press, Oxford

Goulson D, Stout JC, Hawson SA, Allen JA (1998) Floral display size

in comfrey, Symphytum officinale L. (Boraginaceae): relation-

ships with visitation by three bumblebee species and subsequent

seed set. Oecologia 113:502–508

Grindeland JM, Sletvold N, Ims RA (2005) Effects of floral display

size and plant density on pollinator visitation rate in a natural

population of Digitalis purpurea. Funct Ecol 19:383–390

Harder LD, Johnson SD (2009) Darwin’s beautiful contrivances:

evolutionary and functional evidence for floral adaptation. New

Phytol 183:530–545

Hegland S-J, Totland Ø (2005) Relationships between species’ floral

traits and pollinator visitation in a temperate grassland. Oeco-

logia 145:586–594

Hersch EI, Roy BA (2007) Context-dependent pollinator behavior: an

explanation for patterns of hybridization among three species of

Indian paintbrush. Evolution 61:111–124

Jakobsson A, Padron B, Traveset A (2007) Pollen transfer from

invasive Carpobrotus spp. to natives—a study of pollinator

behaviour and reproduction success. Biol Conserv 141:

136–145

Jennersten O (1988) Pollination in Dianthus deltoides (Caryophyll-

aceae): effects of habitat fragmentation on visitation and seed

set. Conserv Biol 2:359–366

Jennersten O, Nilsson SG (1993) Insect flower visitation frequency

and seed production in relation to patch size of Viscaria vulgaris(Caryophyllaceae). Oikos 68:283–292

Kalisz S, Vogler DW, Hanley KM (2004) Context-dependent

autonomous self-fertilization yields reproductive assurance and

mixed mating. Nature 430:884–887

Karoly K (1992) Pollinator limitation in the facultatively autogamous

annual, Lupinus nanus (Leguminosae). Am J Bot 79:49–56

Karron JD, Thumser NN, Tucker R, Hessenauer AJ (1995) The

influence of population density in outcrossing rates in Mimulusringens. Heredity 75:175–180

Klinkhamer PGL, de Jong TJ (1990) Effects of plant size, plant

density and sex differential nectar reward on pollinator visitation

in the protandrous Echium vulgare (Boraginaceae). Oikos

57:399–405

Klinkhamer PGL, de Jong TJ (1993) Atractiveness to pollinators—a

plants dilemma. Oikos 66:180–184

Kudo G (1993) Relationship between flowering time and fruit-set of

the entomophilous alpine shrub, Rhododendron aureum (Erica-

ceae), inhabiting snow patches. Am J Bot 80:1300–1304

Kunin WE (1993) Sex and the single mustard: population density and

pollinator behaviour effects on seed-set. Ecology 74:2145–2160

Kwak MM, Bergman P (1996) Early flowers of Bartsia alpina(Scrophulariaceae) and the visitation by bumblebees. Acta Bot

Neerl 45:355–366

Larson BMH, Barrett SCH (2000) A comparative analysis of pollen

limitation in flowering plants. Biol J Linn Soc 69:503–520

Lazaro A, Totland Ø (2010) Population dependence in the interac-

tions with neighbors for pollination: a field experiment with

Taraxacum officinale. Am J Bot 97:760–769

Lazaro A, Lundgren R, Totland Ø (2009) Co-flowering neighbors

influence the diversity and identity of pollinator groups visiting

plant species. Oikos 118:691–702

Lazaro A, Nielsen A, Totland Ø (2010) Factors related to the inter-

annual variation in plants’ pollination generalization levels

within a community. Oikos 119:825–834

Makino TT, Ohashi K, Sakai S (2007) How do floral display size and

the density of surrounding flowers influence the likelihood of

bumblebee revisitation to a plant? Funct Ecol 21:87–95

McCall C, Primack RB (1992) Influence of flower characteristics,

weather, time of the day, and season on insect visitation rates in

three plant communities. Am J Bot 79:434–442

Mitchell RJ, Karron JD, Holmquist KG, Bell JM (2004) The influence

of Mimulus ringens floral display size on pollinator visitation

patterns. Funct Ecol 18:116–124

Moeller DA (2004) Facilitative interactions among plants via shared

pollinators. Ecology 85:3289–3301

Molano-Flores B, Hendrix SD, Heard SB (1999) The effect of

population size on stigma pollen load, fruit set, and seed set in

Allium stellatum Ker. (Liliaceae). Int J Plant Sci 160:753–757

Mossberg B, Stenberg L (2007) Gyldendals Store Nordiske Flora.

Gyldendal Norsk Forlag, Oslo

Ohashi K, Yahara T (2002) Visit larger displays but probe propor-

tionally fewer flowers: counterintuitive behaviour of nectar-

collecting bumble bees achieves an ideal free distribution. Funct

Ecol 16:492–503

Pauw A (2007) Collapse of a pollination web in small conservation

areas. Ecology 88:1759–1769

Price MV, Waser NM, Irwin RE, Campbell DR, Brody AK (2005)

Temporal and spatial variation in pollination of a montane herb:

a seven-year study. Ecology 86:2106–2116

R Development Core Team (2008) R: A language and environment

for statistical computing. R Foundation for Statistical Comput-

ing, Vienna. ISBN 3-900051-07-0, URL http://www.R-

project.org

Oecologia

123

Rathcke BJ (1983) Competition and facilitation among plants for

pollination. In: Real L (ed) Pollination biology. Academic Press,

Orlando, pp 305–329

Robertson AW, Macnair MR (1995) The effects of floral display size

on pollinator service to individual flowers of Myosotis and

Mimulus. Oikos 72:106–114

Schemske DW (1981) Floral convergence and pollinator sharing in

two bee-pollinated tropical herbs. Ecology 62:946–954

Schmitt J (1983) Flowering plant density and pollinator visitation in

Senecio. Oecologia 60:97–102

Stanton M, Young HJ, Ellstrand NC, Clegg JM (1991) Consequences

of floral variation for male and female reproduction in exper-

imental populations of wild radish, Raphanus sativus L.

Evolution 45:268–280

Stephens DW, Krebs JR (1986) Foraging theory. Princeton University

Press, Princeton

Thompson JD (2001) How do visitation patterns vary among

pollinators in relation to floral display and floral design in a

generalist pollination system? Oecologia 126:386–394

Totland Ø (1993) Pollination in alpine Norway: flowering phenology,

insect visitors, and visitation rates in two plant communities. Can

J Bot 71:1072–1079

Totland Ø (2001) Environment-dependent pollen limitation and

selection on floral traits in an alpine species. Ecology

82:2233–2244

Wissman J (2006) Reproduction and population structure of Primulaveris L. Primulaceae) in semi-natural grasslands—a landscape

and management perspective. Doctoral dissertation, SLU

Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM (2009)

Mixed effects models and extensions in ecology with R.

Springer, New York

Oecologia

123