Interactions among nectar robbing, ﬂoral herbivory, and ant
protection in Linaria vulgaris
Daniel A. Newman and James D. Thomson
Newman, D. A. and Thomson, J. D. 2005. Interactions among nectar robbing, floral
herbivory, and ant protection in Linaria vulgaris./Oikos 110: 497 /506.
Nectar robbers are often assumed to be plant antagonists; however, empirical data
show that the impacts of these animals range from negative to positive depending on
the system and ecological conditions. We experimentally evaluated the combined effects
of nectar robbing and ant visitation on three indices of reproductive fitness in Linaria
vulgaris in the Colorado Rocky Mountains, via indirect effects on flower- and seed-
eating beetles (Brachypterolus pulicarius and Gymnaetron antirrhinni ). Nectar robbing
Bombus occidentalis leave holes in the nectar spurs, effectively creating ‘‘extra-floral
nectaries’’ that attract ants. In a paired-plant experiment, ants were significantly more
abundant on robbed than on unrobbed plants. Manipulation of ant access and nectar
robbing showed that ant exclusion increased beetle attack and decreased female fitness.
There was a significant ant-by-robbing interaction on flower damage. Patterns in the
other two indices were suggestive of ant-by-robbing interactions, but these were not
statistically significant. We also found correlations between spider occupancy on some
plants and the mean number of ants (marginally negative) or beetles (significantly
positive). Although the effect we report in this study may be highly dependent on
spatial and temporal distributions of several interacting species, we discuss its potential
role in mitigating the costs of floral parasitism, and its importance to the study of
nectar robbing in general.
D. A. Newman and J. D. Thomson, Dept of Zoology, Univ. of Toronto, 25 Harbord
Street, Toronto, Ontario, Canada, M5S 3G5 (firstname.lastname@example.org) and Rocky
Mountain Biological Laboratory, Crested Butte, CO 81224, USA.
Researchers of plant/animal interactions have typi-
cally focused on pairwise associations in a single kind
of relationship (e.g. pollination, herbivory, seed-dis-
persal), despite the actual complexity of most systems,
which include several kinds of interaction at once.
More recently, a growing number of papers have
explicitly recognized that an organism’s fitness is the
net result of all its interactions with other organisms
(Herrera 2000, Mothershead and Marquis 2000, Ba-
cher and Friedli 2002, Ehrle´n 2002, Herrera et al.
2002, Bronstein et al. 2003, Stanton 2003, Cariveau
et al. 2004). These studies show that conclusions from
traditional two-species studies can change when the
direct and indirect influences of other species are
considered (Price et al. 1986, Yodzis 1988, Strauss
1991, Wootton 1994, 2002, Menge 1995, Krupnick
et al. 1999, Herrera 2000, Ehrle´n 2002, Herrera et al.
2002, Bronstein et al. 2003).
Pollination biologists have tended to focus on the
direct mutualistic relationship between flowers and the
animals that visit them for pollen and nectar. Recently,
however, this view of pollination relationships has
expanded to include the influence of other species on
animal-mediated plant reproduction. Newer studies
highlight the shortcomings of considering pollination
separately from other types of interaction.
Nectar robbers, because their fitness effects on plants
are generally mediated through changes in the behaviour
Accepted 14 February 2005
Copyright #OIKOS 2005
OIKOS 110: 497 /506, 2005
OIKOS 110:3 (2005) 497
of a third species (the pollinator), have typically been
studied in the context of three-way interactions (robber/
pollinator/plant; Roubik 1982, Roubik et al. 1985, Irwin
and Brody 1999, 2000, Navarro 2000, 2001, Maloof
2001). Nectar robbers bite holes into flowers to obtain
nectar, usually without effecting pollination. Although
they are typically viewed as detrimental to the plants
they attack (but see Zimmerman and Cook 1985,
Navarro 2000, Maloof 2001), and to the pollinators
that they compete with (Mainero and del Rio 1985,
Irwin and Brody 1998), nectar robbers may also
indirectly benefit other animals (including corruptible
pollinators); the holes they bite make nectar more
accessible to others, who become secondary robbers
(Inouye 1980). For example, Roubik et al. (1985) found
that some avian visitors to Quassia amara were limited to
flowers that had previously been robbed. In Linaria
vulgaris in the United Kingdom, some short-tongued
bumblebees collect nectar only through the holes left by
the hole-biting Bombus terrestris (Stout et al. 2000).
Intuition suggests that these additional visitors will have
further negative effects on plant fitness, either through
pollinator avoidance of nectar-depleted flowers, or
through defection of legitimate pollinators to secondary
Some secondarily robbing visitors, however, may
offer benefits to the plants that offset the costs of
robbing. In 2002, we noticed ants feeding on Linaria
vulgaris nectar from holes created by nectar-robbing
bumblebees. This observation led us to hypothesize
that nectar robbing might have positive effects on
L. vulgaris fitness, if ants defended it from its
antagonists. Such defence is a reasonable expectation.
Ant/plant protection mutualisms are widespread,
often mediated by extrafloral nectar (reviewed by
Agrawal and Rutter 1998, Bronstein 1998). This
evolutionary pattern indicates that a plant may
indeed benefit by providing small amounts of sugar
to ants. Thus, nectar made available through holes
made by robbers, might recruit ants as plant
protectors, just as extrafloral nectar does in other
Not all ant/plant associations are beneficial for both
partners; in fact, ants adversely affect some plants by
stealing nectar or damaging flowers (Galen 1983, 1999,
Norment 1988), by eating plant tissues, or by protecting
herbivores (Huxley 1991, Beattie and Hughes 2002). In
some cases, specialized plant traits such as extrafloral
nectaries or food bodies strongly suggest that ants have a
net beneficial effect on the plants. In the absence of such
traits, it is difficult to predict how ants will affect the
plants they visit.
A thorough knowledge of the natural history of
complex systems is required to make relevant predictions
about indirect effects (Menge 1995, Raimondi et al.
2000). In this study, we observed the interactions among
nectar robbing, ants, flower- and seed-eating beetles
(hereafter called ‘‘herbivores’’), and L. vulgaris, a plant
without extrafloral nectaries or other ant-rewarding
traits. We measured the effects of ant recruitment
to robbed plants on several reproductive fitness compo-
nents in L. vulgaris. We hypothesized that nectar robbers
can act as interaction modifiers (sensu Wootton 1994)
that indirectly change the behaviour of ants and
consequently, the abundance or impact of herbivores.
By depositing droplets of sugar solution on experimental
plants, Bentley (1976) demonstrated opportunistic
ant protection of plants that bear no ant-rewarding
traits. Based on her results and our observations of ants
on robbed plants, we posed the following questions:
(1) do ants recruit to robbed flowers to obtain nectar?
And (2) do plants exposed to ants suffer less from
attacks by antagonistic beetles? We predicted that
robbed plants exposed to ants benefit from reductions
Study site and organisms
The study was performed at the Rocky Mountain
Biological Laboratory (RMBL) at Gothic, Colorado
(106859?15ƒN; 38857?30ƒW; 2900 W; 2900 m), between
July 7 and 24, 2003. The area comprises open alpine
meadow, aspen woodland and spruce forest, and human
habitation. Weather during the entire experimental
period was remarkably consistent: no rain, and clouds,
when they occurred, only in the afternoon.
Linaria vulgaris Mill. (Scrophulariaceae) is a long-
lived, perennial, clonal herb introduced to eastern North
America from Eurasia in the 1700s. It has since become a
noxious weed invading pastures and other agricultural
lands, especially in western North America (Saner et al.
1995). Ramets bear several racemose inflorescences with
numerous yellow flowers (mean9
/SD number of open
flowers during daily surveys of this population: 9.459
3.64; n/155; range /1/19) that present nectar in the
end of long spurs (mean9
/SD in this population: 13.379/
3.04 mm; n/673). The flowers’ corolla lips are closed
and thus limit nectar access to insects that are strong
enough to pry them open (generally bumblebees) or,
rarely, animals whose mouthparts are long and thin
enough to fit between them (hummingbirds and hawk-
moths). Bumblebees effect pollination by contacting the
sexual organs on the roof of the flower. Other visitors,
such as very small bees, beetles and ants, can also fit or
force themselves between the corolla lips (D. A. New-
man, pers. obs.). Linaria vulgaris reproduces clonally by
aggressive rhizomatous growth and sexually by prolific
Three studies have investigated the pollinator-
mediated effects of nectar robbers on female fitness
498 OIKOS 110:3 (2005)
in L. vulgaris (Stout et al. 2000, Irwin and Maloof
2002, Nepi et al. 2002); none found significant
reductions in seed set, presumably because so few
legitimate visits are required to fertilize all available
ovules (Arnold 1982). At the RMBL, seed set in L.
vulgaris is not pollination limited, even in heavily
robbed patches (Irwin and Maloof 2002, R. E. Irwin,
pers. comm.). Attack by two beetle species, Brachyp-
terolus pulicarius (L.), a flower-eating nitidulid, and
Gymnaetron antirrhinni Paykull, a seed-eating weevil,
however, is known to strongly affect female fitness
(Smith 1959, Harris 1961, McClay 1992, Saner et al.
1995). The two beetles may compete for resources and
may displace one another in some cases (Harris 1961).
Of the two beetles, B. pulicarius is more common at
the RMBL (D. A. Newman, pers. obs.). Adults of this
species emerge in spring and feed upon young L.
vulgaris stems; when flowers begin to form, they
oviposit in the buds, where larvae subsequently feed
on the anthers and ovaries. Older larvae are also
known to eat seeds (Harris 1961). The larvae are able
to move from flower to flower and can reduce seed set
by 75% and seed weight by 60% (McClay 1992).
Gymnaetron antirrhinni adults emerge in spring and
eat young stems, then lay eggs in the ovaries during
flowering. In the developing fruit, the larvae feed on
seeds, then pupate, eclose, and overwinter as adults.
This species can also reduce seed set by more than
50% (Harris 1961).
We worked in a 10/6 m plot that contained an early
blooming patch of L. vulgaris plants. Because there were
no nectar robbers foraging at the time of the experiment,
we simulated hole-biting by piercing the middle of the
spur with fine-nosed forceps whose tips had been bent
inward. The location and size of the artificial robber
holes matched those of real robbers observed in 2002 at
the RMBL. Nectar did not leak out of the holes, but
could be reached by ants that could fit their mouthparts,
heads, or whole bodies through them. To control for the
disturbance that piercing may have caused (e.g. frighten-
ing ants and beetles), we also snapped the tweezers shut
over the spurs of all flowers in the unrobbed treatments,
but without making holes. We pierced or sham-pierced
all flowers on the experimental plants, including all
flower buds that were developed enough to contain
Ant recruitment to robbed and unrobbed plants
Between July 7 and 12, in an area with high ant
activity, we haphazardly chose five pairs of racemes
with a large number of open flowers (mean9
/3.88) and pierced all the flowers on one
inflorescence of each pair. Paired ramets most likely
belonged to the same genet (we selected pairs that
appeared to grow out of the same root system), and
were near neighbours (
/20 cm). Three times a day
(morning, noon, and late afternoon) for a total of 11
sampling episodes, we scored the ant activity on each
plant, as follows: number of ants of each type, behaviour
(patrolling plant, drinking nectar), and location on
the ramet (lower leaves, inflorescence, within or on
flowers). To avoid problems associated with the inter-
dependence of ant sightings over subsequent sampling
periods, we averaged data for each plant over all
Effects of ants and nectar robbing on flower and
seed predators and seed set
On July 7 where enough racemes approaching flowering
could be found, we chose four neighbouring ramets
/20 cm apart) to represent four treatments in a full
factorial design. These were: (1) ant-excluded, pierced
spurs; (2) ant-excluded, unpierced spurs; (3) ant-access,
pierced; and (4) ant-access, unpierced (hereafter, treat-
ment names for all experiments will be capitalized). We
pierced or sham-pierced every flower on all experimental
ramets. In all, we treated 160 ramets (n
excluded ants by removing the leaves from each plant’s
/15 lower nodes and applying Tanglefoot†to the stem;
on plants where ants were allowed, we removed leaves
but did not apply Tanglefoot. We also removed any
vegetation that touched or could touch the experimental
plants to prevent ants from accessing them. The piercing
treatment, including the sham piercing of flowers on the
unpierced plants, was administered as described above.
Ideally, we would have hand-pollinated the flowers to
remove any pollinator preference for one treatment over
another, but the disturbance that this would have caused
for the insect community inhabiting the plants would
have greatly affected the results of the study; hand-
pollinating L. vulgaris closed flowers is invasive enough
to cause the beetles to leave the plants (D. A. Newman,
pers. obs.). We noticed no differences in either the
behaviour or the frequency of bumblebee visitors among
Between July 8 and 24, we surveyed the experimental
plants, once each morning (
/09:00 to /11:30 h) and
once each afternoon (
/14:00 to /16:30 h). During
these surveys, we pierced or sham-pierced new flowers,
counted the number of beetles and ants visible on the
plants, and scored their activities. In addition, we noted
the presence of other animals on the plants (Results;
Animals on Linaria vulgaris). Since both the nitidulids
and the weevils sometimes hide inside flowers, their
abundance on plants was probably underestimated. We
OIKOS 110:3 (2005) 499
also noted any behavioural interactions among ants,
beetles, and any other animals.
Effect of beetles and ants on plant fitness indices
Every day, we counted the number of mature flowers,
maturing flowers (large buds with developed spurs), and
flower buds on every plant. We also collected wilted
flowers before abscision, and inspected them for beetle
frass and larvae. We collected the mature fruit on August
16, counted them, then removed the seeds to count and
weigh them. We recorded the presence of larval and
mature beetles in the fruit.
We measured eight variables (Table 1). Most represent
components of female fitness, although male fitness may
also be inferred from some variables (e.g. damaged or
frass-soiled flowers may be less attractive to pollinators,
and the quality or quantity of pollen in flowers with
many larvae might be reduced; Quesada et al. 1995,
Strauss 1997). We then created three fitness ‘‘indices’’, an
herbivore-damage index (mean% damaged flowers), a
beetle-attack index [(% frass-soiled flowers)
number of beetle larvae/flower)], and a female fitness
index [(number of fruits/flower)
/(number of seeds/
/(mean seed weight)] by combining variables
shown in Table 1.
Plant, ant, and beetle densities appeared to
vary greatly within the experimental plot. To reduce
random effects due to this variability, plants were
grouped in fours with their neighbours (i.e. each
group included one plant of each treatment); within
groups, the mean value was subtracted from each data
point (Quinn and Keough 2002). This standardization
had the additional benefit of normalizing residuals and
homogenizing variances. The effects of ant-access, pier-
cing and ant-access
/piercing interactions on the fitness
indices were tested with two-way ANOVAs in SPSS 10.0
(SPSS Inc. 1989 /9999). By the end of the study, some
individual plants had died, been severely damaged by
rodents, or aborted all of their flowers. Sample sizes vary
because some data were collected earlier in the season
when fewer plants were damaged. Only groups
represented by plants of all four treatments were
included in the analyses. Unstandardized means and
standard errors for the eight variables are presented in
Animals on Linaria vulgaris
Linaria vulgaris plants were host to a wide variety of
arthropods ranging from pollinators to herbivores to
ambush predators that used them as habitat. Pollinators
included four bumblebees (Bombus bifarius,B.
flavifrons,B. appositus and B. californicus ), Anthophora
furcata-terminalis, and, potentially, a small number of
unidentified halictid bees.
Both G. antirrhinni and B. pulicarius frequently flew
from plant to plant. Both were often observed mating on
the flowers throughout the study and in the weeks
following it. Brachypterolus pulicarius larvae typically
resided in the flower vestibule (the cavity into which a
pollinator’s head and thorax fit while it reaches for
nectar), although we also often found them in the spurs.
Flowers infested with the beetle larvae were often soiled
with frass and, in several cases, had shrunken or
otherwise damaged anthers.
Three species of ants were frequently observed on L.
vulgaris. These were Formica lasioides Emery, F. fusca
sspp. and Tapinoma spp. Formica fusca sspp. and
Tapinoma spp. were the most common. Tapinoma, the
smallest of the ants observed (mean9
mm), can squeeze between the closed lips of L. vulgaris
and was often seen drinking nectar within the spur. The
two larger species (mean9
/SD: F. fusca sspp./5.089/
0.576 mm and F. lasioides/3.879/0.291 mm) occasion-
ally entered flowers that were open either due to
malformation or wilting. Ants were divided into large
(both Formica species) and small (Tapinoma spp.)
categories for some analyses because large and small
ants can exert predation pressures on different life
history stages of their prey (Cushman and Addicott
1991). Tapinoma spp. could fit through the holes in the
spurs and often imbibed nectar from within the flowers;
the larger ants typically drank nectar at the opening of
the hole or pushed their heads within the spur. We also
observed a small number of ants biting the spur tissue
around the piercing, apparently enlarging the hole. Ants
of a fourth taxon, F. fusca neorufibarbis Emery,
although abundant in the study plot and known to
thieve nectar from other systems in the area (Galen 1983,
Norment 1988), never visited L. vulgaris plants.
Ants were patchy in their distribution and were only
rarely seen in the vicinity of some of the plants; ants were
never observed on 11 groups of plants. Conversely, some
plants, especially those growing on a gravel slope on the
west side of the plot, were consistently visited by ants.
Table 1. List of variables measured for each plant. Asterisks
denote variables that measure ‘‘negative’’ ﬁtness components.
Number of flowers per plant
Number of damaged flowers*
Number of flowers with beetle frass*
Number of beetle larvae per plant*
Number of enlarged fruit
Number of unenlarged (failed) fruit*
Total number of seeds
Total seed weight
500 OIKOS 110:3 (2005)
Ants visiting the plants typically patrolled the stem and
flowers and occasionally chewed on floral tissues.
Several plants also hosted spiders (salticids, thomisids,
and theridiids), which were occasionally observed killing
ants and, once, a nitidulid. Spiders typically remained on
the same plants over the survey period.
Observations of behavioural and other species
Although ants and beetles often inhabited the same
plant at the same time, we rarely observed interactions
between them during the five-minute surveys. In those
cases, however, we did see ants attacking and, in two
cases, driving away nitidulids or weevils. Ants also
reacted aggressively to experimental tools (e.g. tweezers,
human hands) and to visiting bumblebees, in one case
driving a Bombus bifarius worker away from the plant.
Nitidulids and weevils rarely interacted because they
were rarely found on the same plants; over the course of
the survey period we saw plants with at least one beetle
of either species 627 times, with nitidulids alone 571
times, and with weevils alone 20 times. We observed
plants with both only 16 times, constituting a significant
negative association (x
/5.268, df /1, p/0.022). When
we found both species together, they seldom reacted to
each other’s presence. In one case, however, a weevil
charged a nitidulid, driving it off the plant.
Spiders occasionally attacked and killed ants, and the
webs that enmeshed some inflorescences often contained
several dead ants. One crab spider was also observed
holding a dead nitidulid. Crab spiders were, like ants,
quite aggressive. They attacked almost any object
approaching the plant, including human fingers and
Within the fruit, which we dissected in October 2003,
we found in addition to larval and adult weevils a small
number of adult parasitic wasps of the genus Pteroma-
lus, a group known to attack cucurlionid beetles
(S. Libenson, pers. comm.).
Ant recruitment to robbed versus unrobbed plants
We observed more ants per survey on robbed
plants than on unrobbed plants. For the five plant
pairs (data from the 11 sampling epidsodes are pooled
for each plant), the number of ants was higher on
the robbed individuals than on their unrobbed neigh-
/SD: 3.6369/4.076 and 0.9829/0.360,
respectively; Wilcoxon matched pairs test: Z
/0.043, n/5). There were significantly more large
ants on the robbed plants than on the unrobbed plants
/0.342 and 0.7099/0.088, respectively; Wilcoxon
matched pairs test: Z
/2.023, p/0.043, n /5) and
no difference in the number of small ants among
treatments. During the surveys, there were more
instances of ants feeding (robbing or forcing
their ways into flowers) on robbed plants, whereas
ants on unrobbed plants were more likely to be
patrolling or sitting motionless (x
/27.092, df /1,
Ants and herbivores on Linaria vulgaris
There was no difference in mean (9/SE) ant abundance
between the two treatments that allowed ant access
/0.3019/0.106, ants-unpierced /0.3119/
0.074; Mann/Whitney U test: z/1.032, p/0.302, n /
38); there was also no difference in recruitment of small
or large ants to either treatment (x
/2.61, df /1, p/
0.106). There was, however, a difference in activity type
among the ant-access treatments: on plants whose
flowers were pierced, significantly more ants were seen
on or in flowers, than in plants whose flowers were intact
/27.71, df /1, pB/0.00001).
Ant exclusion had a marginal negative effect on
/SE) number of beetles per plant per survey
/0.2929/0.028, ant-excluded /0.4319/0.054,
Mann/Whitney U test: z/1.83; p/0.067; n/155
Fig. 1), the hole treatment had no effect (holes
0.4149/0.054, no holes /0.3109/0.030; Mann /Whitney
Ants No ants
Ant access treatment
Fig. 2. Effects of the ant and piercing treatments on the
herbivore damage index (% damaged ﬂowers per plant per
survey). Circles are means9
Ants No ants
Ant access treatment
Fig. 1. Effect of ant access and spur piercing on beetle
abundance on Linaria vulgaris ramets.
OIKOS 110:3 (2005) 501
U test: z/0.992, p /0.321, n/155). However, ant-
excluded, pierced plants had significantly more beetles
than any of the other treatments (Wilcoxon matched
pairs tests; Z
/2.482, pB/0.020, Fig. 2).
The presence of spiders was not influenced by the
/0.857; df /3; p/0.836). However, spi-
der occupancy (spiders per plant per observation) on
plants was positively correlated with beetle abundance
/0.280, p /0.0004, n/155), and marginally nega-
tively correlated with ant abundance (r
Effects of beetles and ants on plant fitness indices
Ant exclusion had significant effects on the beetle
attack and female fitness indices, while the piercing
/piercing interaction did not (Table 2).
Neither ant-access nor piercing significantly affected
the flower-damage index, but the ants
action was significant; holes in flowers appear to be
detrimental when ants are excluded but positive
when they are not (Fig. 2). Although not significant,
this pattern is also found in the other two indices
(Fig. 3, 4).
Mean observed ant abundance on plants was signifi-
cantly correlated with more than half of the measured
variables (Table 3); significant correlations with ‘‘nega-
tive fitness measures’’ were negative, and significant
correlations with ‘‘positive fitness measures’’ were all
positive. In addition, all correlation coefficients for the
non-significant correlations were consistent in direction
with our predictions about the positive effects of ants on
plant fitness. Only number of frass-soiled flowers was
significantly (and positively) correlated with mean beetle
abundance (Table 3). Spider occurrence on plants was
not correlated with any of the variables (r
B/0.15; p /
Because they divert nectar from legitimate pollinators,
nectar robbers can be seen as disrupting a mutualistic
relationship between Linaria vulgaris and its pollinators.
Species that disrupt mutualisms (aprovechados, Mainero
and del Rio, 1985) are generally assumed to reduce the
fitness of at least one of the two mutualistic partners.
Although both negative and positive effects on plant
fitness have been attributed to nectar robbers (reviewed
by Maloof and Inouye 2000, Irwin et al. 2001,), most
workers still consider robbers to be parasites of plant/
pollinator mutualisms, and implicitly treat them as plant
antagonists (Proctor et al. 1996, Traveset et al. 1998,
Irwin and Brody 1999, 2000, Stout et al. 2000, Bronstein
2001, Lara and Ornelas 2001, Yu 2001, Anderson and
Midgley 2002, Irwin 2003, Stanton 2003). Here we show
that flower piercing may indirectly benefit a plant, L.
vulgaris, by attracting secondary robbers (ants) that also
reduce the impact of herbivores. Primary nectar robbers,
by making holes in the nectar spurs, effectively create a
plant trait that creates or strengthens a protective
ant/plant interaction. Ants that normally explore the
plants in low numbers may, upon finding access to
nectar through holes in the spurs, recruit to robbed
plants in much the same way as they would to plants
bearing extra-floral nectaries. It is also plausible that
Table 2. Results of two-way ANOVAs on ﬁtness indices in
Linaria vulgaris. Sample sizes varied for analyses of different
variables because plants did not always survive long enough to
produce mature fruit.
Index Treatment df
ants 1, 128 2.647 0.106
piercing 1.976 0.162
/piercing 4.300 0.040
Beetle attack ants 1, 116 10.28 0.002
piercing 0.953 0.953
/piercing 0.648 0.648
ants 1, 92 9.702 0.002
piercing 0.397 0.530
Female fitness ants
/piercing 2.411 0.124
Ants No ants
Ant access treatment
Fig. 3. Effects of the ant and piercing treatments on the beetle
attack index ((proportion of ﬂowers soiled with frass per
/(mean number of beetle larvae per ﬂower)). Circles
Ants No ants
Ant access treatment
Fig. 4. Effects of the ant and piercing treatments on the female
ﬁtness index ((number of fruits/ﬂower)
/(number of seeds/
/(mean seed weight)). Circles are means9/SE.
502 OIKOS 110:3 (2005)
ants can detect chemical indications of nectar robbing
activity (i.e. the smell of nectar or other volatiles released
by the corolla tissues). Our results also suggest that
pierced flowers attract the beetles (Fig. 1, 2), but it
appears in this case that the benefit of recruited ants
exceeds the costs of recruited herbivores. In this study,
ant access positively affected plant fitness components.
It remains unclear whether ants act upon beetles
primarily through consumptive or behavioural effects
(Rudgers et al. 2003), and whether they act more upon
adult beetles or larvae that can presumably be
accessed through holes in the spurs (D. A. Newman,
To our knowledge, this is the first study to show that
nectar robbing can have indirect effects on plant fitness
through mechanisms other than pollinator behaviour.
Although the ant exclusion study did not reveal an effect
of the flower piercing (artificial robbing) treatment on
any fitness component, the ant recruitment study,
conducted in a small subset of the plot with consistently
high ant activity, did. Ants on pierced plants also
behaved differently than those on unpierced plants;
they were significantly more often feeding on nectar or
patrolling flowers, where aggressive interactions with the
herbivorous beetles were most likely to occur. In addi-
tion, the significant ant
/piercing interaction effect
found for the herbivore damage index, and the non-
significant patterns that suggest such an interaction in
the other two indices, support the prediction that holes
may have effects contingent upon ants for at least some
components of reproductive fitness. Our inability to
show significant effects of the robbing treatment in the
exclusion study is most likely due to spatial variation in
ant recruitment within the study plot (see Bronstein 1998
on intraspecific variation in ant/plant protection mutu-
alisms). Ant abundances are known to vary drastically
within very small areas, depending in large part on the
proximity of colonies (Cushman and Addicott 1991).
Although the interactions we demonstrate may be strong
only in some patches, they could still be ecologically
important. Indeed, even if the average effect of ants and
robbers (and spiders, for that matter) on the success of L.
vulgaris is very small, as long as it is sometimes locally
strong, as we have shown in this study, it could amount
to more than a simple ecological curiosity (Berlow 1999).
Indeed, if plants located close to ant colonies (which are
often long-lived, and would thus be associated with the
same L. vulgaris individuals for several years; Cushman
and Addicott 1991) are consistently enjoying greater
reproductive success than plants in ant-poor areas which
may allocate more resources to vegetative growth than to
sexual reproduction, a differential contribution of the
ant-attended plants to future generations could be
Nectar robbers can have indirect and counterintuitive
effects on reproductive success in L. vulgaris. Their
effects on pollinator behaviour and the resulting bene-
ficial impacts on plant fitness have been well studied
(reviewed by Maloof and Inouye 2000); here, we show
that taxa unrelated to pollination may also benefit
nectar-robbed plants. One of the important questions
in community ecology is how indirect an effect can be
without becoming completely negligible. Theorists have
proposed that indirect effects should be attenuated when
many species interact due to dilution of interaction
strength, resulting from the magnification of noise as the
number of links in the assemblage increases (Strauss
1991, Wootton 1994, Williams et al. 2002). However,
theoretical and empirical studies have also shown that
indirect effects may be as strong as, or stronger than,
direct effects (Wootton 1994, Menge 1995, Berlow 1999,
Williams et al. 2002). In this study, simulated nectar
robbing, whose direct and pollinator-mediated indirect
effects on female fitness are weak or non-existent, had
highly indirect but significant effects on components
of reproductive success. It is important to note that,
unlike real nectar robbers, simulated robbing did not
remove nectar from the spurs. However, we found that
primary and secondary robbers did not always remove
all the nectar from L. vulgaris at the RMBL, and that
nectar production did not change in robbed flowers (D.
A. Newman, pers. obs.). In addition, ants were even
attracted to minute quantities of floral rewards,
and would therefore still recruit to robbed plants whose
holes allow access to traces of nectar; for example, we
often observed ants feeding on immeasurably small
amounts of sugar around the ovaries of recently abscised
It is possible that beneficial indirect effects of ants
on L. vulgaris were tempered somewhat by negative
direct effects of their foraging activities. Indeed, we
observed ants chewing on floral tissues in both ant-
access treatments; this may have detrimental impacts on
flower attractiveness, flower longevity and fruit success
(Krupnick and Weis 1998, Krupnick et al. 1999, Utelli
and Roy 2001). Such conditionality in the consequences
Table 3. Spearman correlations between mean ant and beetle
abundances and plant ﬁtness variables.
/0.126 0.145 0.071 0.412
Flowers with frass*
/0.209 0.037 0.224 0.025
Number of larvae*
/0.294 0.005 0.120 0.264
Number of fruit 0.132 0.138 0.129 0.147
Number of failed
/0.040 0.655 0.136 0.126
Number of seeds 0.192 0.030 0.067 0.452
Mean seed weight 0.302 0.003
OIKOS 110:3 (2005) 503
of species interactions is well known (Cushman and
Addicott 1991, Bronstein et al. 2003, reviewed by
Thompson 1988 and by Bronstein 1998) and could,
depending on ecological circumstances, result in inter-
actions that span from mutualism to commensalism to
In addition to the main effects, we found suggestive
evidence of an additional guild that, through yet another
indirect step, may modify the observed robber/ant /
beetle interactions. Plants that often hosted spiders
hosted more beetles and almost significantly fewer ants
than plants that hosted none or few. We saw spiders kill
ants by ambush or in their webs. Gastreich (1999) found
that theridiid spiders kill Pheidole bicornis ants visiting
Piper obliquum, a plant with which they have a
facultative protection mutualism; the resulting avoidance
of plants with webs by ants caused an increase in
herbivory. Due to the small number of plants that hosted
spiders in this study, however, we could detect no
significant correlations between spider occurrence and
any of the fitness variables; it is likely that the effect of
spiders in this system is highly incidental and very
The effect we demonstrate in this study is almost
certainly dependent on the floral morphology of plants
that, like L. vulgaris, can deter nectar-seeking ants.
Without closed corollas, which provide an effective
barrier between ants and nectar, robbers would not
modify the physical availability of ‘‘reward’’ and would
therefore contribute no additional ant protection. The
plant’s natural history is also probably important. As
discussed earlier, researchers have found that nectar
robbing has no significant impact on the female fitness
of L. vulgaris (Stout et al. 2000, Irwin and Maloof 2002,
Nepi et al. 2002); in a species that suffered reductions in
pollination due to nectar robbers, it is doubtful that ants
attracted to robbed plants could counter the negative
effects of pollen limitation.
One of the factors that has kept nectar robbing,
despite its ubiquity and taxonomic breadth, from
appreciation as an important phenomenon in plant
evolutionary ecology (Irwin et al. 2001, Ehrle´n 2002) is
the lack of generality that researchers have found in
the outcomes of plant /pollinator/robber interactions
(Morris 1996, Maloof and Inouye 2000, Agrawal 2001,
Irwin et al. 2001). For this reason, studies have
been aimed primarily at demonstrating whether nectar
robbers are parasites or mutualists, while very few (Lara
and Ornelas 2001) have even implicitly addressed the
potentially more interesting questions about how and
why the magnitude and sign of their effects vary in time,
in space, and among systems (Cushman and Addicott
1991, Irwin and Maloof 2002). In the system described
in this study, we provide one of many potential
mechanisms that may cause the effects of nectar robbing
to vary; researchers interested in this and other ecologi-
cal phenomena may benefit from considering species
that do not appear to interact with their focal species.
For the development of relevant questions and hypoth-
eses in such complex systems, however, it is critical to
base them not only on the rapidly growing body of
theory dealing with indirect effects in ecological com-
munities, but also on intimate knowledge of their natural
Acknowledgements /The authors wish to thank A. A. Agrawal,
K. N. Jones, J. S. Reithel, J. S. Thaler and P. Wilson for useful
discussion, M.-J. Fortin and D. A. Jackson for statistical advice
and I. Billick and S. Libenson for help in identifying ants and
wasps, respectively. E. Ranta provided valuable comments on
the manuscript. Funding to D. A. Newman was provided by the
Natural Sciences and Engineering Research Council of Canada
(NSERC) and the University of Toronto.
Agrawal, A. A. 2001. Nectar, nodules and cheaters. /Trends
Ecol. Evol. 16: 123 /124.
Agrawal, A. A. and Rutter, M. T. 1998. Dynamic anti-herbivore
defense in ant-plants: the role of induced responses. /Oikos
83: 227 /236.
Anderson, B. and Midgley, J. J. 2002. It takes two to tango but
three is a tangle: mutualists and cheaters on the carnivorous
plant Roridula ./Oecologia 132: 369 /373.
Arnold, R. M. 1982. Pollination, predation and seed set in
Linaria vulgaris (Scrophulariaceae). /Am. Midl. Nat. 107:
Bacher, S. and Friedli, J. 2002. Dynamics of a mutualism in a
multi-species context. /Proc. R. Soc. Lond. B 269: 1517 /
Beattie, A. J. and Hughes, L. 2002. Ant /plant interactions.
/In: Herrera, C. M. and Pellmyr, O. (eds), Plant/animal
interactions: an evolutionary approach. Blackwell Science,
pp. 211 /236.
Bentley, B. L. 1976. Plants bearing extraﬂoral nectaries and the
associated ant community: interhabitat differences in the
reduction of herbivore damage. /Ecology 57: 815 /820.
Berlow, E. L. 1999. Strong effects of weak interactions in
ecological communities. /Nature 398: 330 /334.
Bronstein, J. L. 1998. The contribution of ant/plant protection
studies to our understanding of mutualism. /Biotropica 30:
Bronstein, J. L. 2001. The exploitation of mutualisms. /Ecol.
Lett. 4: 277 /287.
Bronstein, J. L., Wilson, W. G. and Morris, W. F. 2003.
Ecological dynamics of mutualist/antagonist communities.
/Am. Nat. 162: S24/S39.
Cariveau, D., Irwin, R. E., Brody, A. K. et al. 2004. Direct and
indirect effects of pollinators and seed predators to selection
on plant and ﬂoral traits. /Oikos 104: 15 /26.
Cushman, J. H. and Addicott, J. F. 1991. Conditional interac-
tions in ant /plant/herbivore mutualisms. /In: Huxley, C.
R. and Cutler, C. F. (eds), Ant /plant interactions. Oxford
Univ. Press, pp. 92 /103.
Ehrle´ n, J. 2002. Assessing the lifetime consequences of plant /
animal interactions for the perennial herb Lathyrus vernus
(Fabaceae). /Persp. Plant. Ecol. Evol. Syst. 5: 145 /163.
Galen, C. 1983. The effects of nectar thieving ants on seedset in
ﬂoral scent morphs of Polemonium viscosum./Oikos 41:
Galen, C. 1999. Flowers and enemies: predation by nectar-
thieving ants in relation to variation in ﬂoral form of an
alpine wildﬂower, Polemonium viscosum./Oikos 85: 426 /
504 OIKOS 110:3 (2005)
Gastreich, K. R. 1999. Trait-mediated indirect effects of a
Theridiid spider on an ant /plant mutualism. /Ecology 80:
Harris, P. 1961. Control of toadﬂax by Brachypterolus pulicarius
(L.) (Coleoptera: Nitidulidae) and Gymnaetron antirrhini
(Payk.) (Coleoptera: Curculionidae) in Canada /Can.
Entomol. 43: 977 /981.
Herrera, C. M. 2000. Measuring the effects of pollinators and
herbivores: evidence for non-additivity in a perennial herb.
/Ecology 81: 2170 /2176.
Herrera, C. M., Medrano, M., Rey, P. J. et al. 2002. Interaction
of pollinators and herbivores on plant ﬁtness suggests a
pathway for correlated evolution of mutualism- and antag-
onism-related traits. /Proc. Natl Acad. Sci. USA 99:
Huxley, C. R. 1991. Ants and plants: a diversity of interactions.
/In: Huxley, C. R. and Cutler, C. F. (eds), Ant/plant
interactions. Oxford Univ. Press, pp. 1/11.
Inouye, D. W. 1980. The terminology of nectar larceny.
/Ecology 61: 1251 /1253.
Irwin, R. E. 2003. Impact of nectar robbing on estimates of
pollen ﬂow: conceptual predictions and empirical outcomes.
/Ecology 84: 485 /495.
Irwin, R. E. and Brody, A. K. 1998. Nectar robbing in
Ipomopsis aggregata: effects on pollinator behavior and
plant ﬁtness. /Oecologia 116: 519/527.
Irwin, R. E. and Brody, A. K. 1999. Nectar-robbing bumble
bees reduce the ﬁtness of Ipomopsis aggregata (Polemonia-
ceae). /Ecology 80: 1703 /1712.
Irwin, R. E. and Brody, A. K. 2000. Consequences of nectar
robbing for realized male function in a hummingbird-
pollinated plant. /Ecology 81: 2637 /2643.
Irwin, R. E. and Maloof, J. E. 2002. Variation in nectar
robbing over time, space, and species. /Oecologia 133:
Irwin, R. E., Brody, A. K and Waser, N. M. 2001. The impact of
ﬂoral larceny on individuals, populations, and communities.
/Oecologia 129: 161 /168.
Krupnick, G. A. and Weis, A. E. 1998. Floral herbivore effect
on the sex expression of an andromonoecious plant,
Isomeris arborea (Capparaceae). /Plant Ecol. 134: 151 /
Krupnick, G. A., Weis, A. E. and Campbell, D. R. 1999. The
consequences of ﬂoral herbivory for pollinator service to
Isomeris arborea ./Ecology 80: 125 /134.
Lara, C. and Ornelas, J. F. 2001. Preferential nectar robbing of
ﬂowers with long corollas: experimental studies of two
hummingbird species visiting three plant species. /Oecolo-
gia 128: 263 /273.
Mainero, J. S. and del Rio, C. M. 1985. Cheating and taking
advantage in mutualistic associations. /In: Boucher, D. H.
(ed.), The biology of mutualism: ecology and evolution.
Croom Helm, pp. 192 /216.
Maloof, J. E. 2001. The effects of a bumble bee nectar robber on
plant reproductive success and pollinator behaviour. /Am.
J. Bot. 88: 1960 /1965.
Maloof, J. E. and Inouye, D. W. 2000. Are nectar robbers
cheaters or mutualists? /Ecology 81: 2651 /2661.
McClay, A. S. 1992. Effects of Brachypterolus pulcarius
(L.) (Coleoptera: Nitidulidae) on ﬂowering and seed pro-
duction of common toadﬂax. /Can. Entomol. 124: 631 /
Menge, B. A. 1995. Indirect effects in marine rocky intertidal
interaction webs: patterns and importance. /Ecol. Monogr.
65: 25 /74.
Morris, W. F. 1996. Mutualism denied? Nectar-robbing bumble
bees do not reduce female or male success in bluebells.
/Ecology 77: 1451 /1462.
Mothershead, K. and Marquis, R. J. 2000. Fitness impacts of
herbivory through indirect effects on plant /pollinator
interactions in Oenothera macrocarpa./Ecology 81: 30 /
Navarro, L. 2000. Pollination ecology of Anthyllis vulneraria
subsp. vulgaris (Fabaceae): nectar robbers as pollinators.
/Am. J. Bot. 87: 980 /985.
Navarro, L. 2001. Reproductive biology and effect of nectar
robbing on fruit production in Macleania bullata (Erica-
ceae). /Plant Ecol. 152: 59/65.
Nepi, M., Pacini, E., Nencini, C. et al. 2002. Variability
in nectar production and composition in Linaria vulgaris
(L.) Mill. (Scrophulariaceae). /Plant Syst. Evol. 238: 109 /
Norment, C. J. 1988. The effect of nectar-thieving ants on the
reproductive success of Frasera speciosa (Gentianaceae).
/Am. Midl. Nat 120: 331 /336.
Price, P. W., Westoby, M., Rice, B. et al. 1986. Parasite
mediation in ecological interactions. /Annu. Rev. Ecol.
Syst. 17: 487 /505.
Proctor, M., Yeo, P. and Lack, A. 1996. The natural history of
pollination. /Timber Press, Portland, Oregon, USA.
Quesada, M., Bollman, K. and Stephenson, A. G. 1995. Leaf
damage decreases pollen production and hinders
pollen performance in Cucurbita texana ./Ecology 76:
Quinn, G. P. and Keough, M. J. 2002. Experimental design and
data analysis for biologists. /Cambridge Univ. Press.
Raimondi, P. T., Forde, S. E., Delph, L. F. et al. 2000. Processes
structuring communities: evidence fro trait-mediated indir-
ect effects through induced polymorphisms. /Oikos 91:
Roubik, D. W. 1982. The ecological impact of nectar-robbing
bees and pollinating humming birds on a tropical shrub.
/Ecology 63: 354 /360. 1982.
Roubik, D. W., Holbrook, N. M. and Parra, G. 1985. Roles
of nectar robbers in reproduction of the tropical
treelet Quassia amara (Simaroubaceae). /Oecologia 66:
Rudgers, J. A., Hodgen, J. G. and White, J. W. 2003. Behavioral
mechanisms underlie an ant-plant mutualism. /Oecologia
135: 51 /59.
Saner, M. A., Clements, D. R., Hall, M. R. et al. 1995. The
biology of Canadian weeds. 105. Linaria vulgaris Mill.
/Can. J. Plant. Sci. 75: 525 /537.
Smith, J. M. 1959. Notes on insects, especially Gymnaetron spp.
(Coleoptera: Curculionidae), associated with toadﬂax, Li-
naria vulgaris Mill. (Scrophulariaceae), in North America.
/Can. Entomol. 91: 116 /121.
Stanton, M. L. 2003. Interacting guilds: moving beyond the
pairwise perspective on mutualisms. /Am. Nat. 162: S10 /
Stout, J. C., Allen, J. A. and Goulson, D. 2000. Nectar robbing,
forager efﬁciency and seed set: bumblebees foraging on the
self-incompatible plant Linaria vulgaris (Scrophulariaceae).
/Acta Oecol. 21: 277 /283.
Strauss, S. Y. 1991. Indirect effects in community ecology: their
deﬁnition, study and importance. /Trends Ecol. Evol. 6:
Strauss, S. Y. 1997. Floral characters link herbivores, pollina-
tors, and plant ﬁtness. /Ecology 78: 1640 /1645.
Thompson, J. N. 1988. Variation in interspeciﬁc interactions.
/Annu. Rev. Ecol. Syst. 19: 65 /87.
Traveset, A., Willson, M. F. and Sabag, C. 1998. Effect of
nectar-robbing birds on fruit set of Fuchsia magellanica in
Tierra Del Fuego: a disrupted mutualism. /Funct. Ecol. 12:
Utelli, A.-B. and Roy, B. A. 2001. Causes and consequences
of ﬂoral damage in Aconitum lycoctonum at high
and low elevations in Switzerland. /Oecologia 127: 266 /
Williams, R. J., Berlow, E. L., Dunne, J. A. et al. 2002. Two
degrees of separation in complex food webs. /Proc. Natl
Acad. Sci. 99: 12913 /12916.
Wootton, J. T. 1994. The nature and consequences of indirect
effects in ecological communities. /Annu. Rev. Ecol. Syst.
25: 443 /466.
OIKOS 110:3 (2005) 505
Wootton, J. T. 2002. Indirect effects in complex ecosystems:
recent progress and future challenges. /J. Sea Res. 48: 157/
Yodzis, P. 1988. The indeterminacy of ecological experiments as
perceived through perturbation experiments. /Ecology 69:
Yu, D. W. 2001. Parasites of mutualisms. /Biol. J. Linn. Soc.
72: 529 /546.
Zimmerman, M. and Cook, S. 1985. Pollinator foraging,
experimental nectar-robbing and plant ﬁtness in Impatiens
capensis./Am. Midl. Nat. 113: 84/91.
Subject Editor:Esa Ranta
Appendix 1. Means (9/SE) for unstandardized variables.
Ants, pierced Ants, unpierced No ants, pierced No ants, unpierced
Mean number of ants per survey (n
/34) 0.155 (0.039) 0.182 (0.038) 0 (0) 0 (0)
Mean number of beetles per survey (n
/34) 0.265 (0.041) 0.257 (0.040) 0.480 (0.078) 0.298 (0.089)
Flower number (n
/34) 12.80 (1.424) 11.93 (1.222) 11.567 (1.053) 12.33 (1.139)
Number of damaged flowers (n
/34) 0.300 (0.063) 0.355 (0.074) 0.563 (0.105) 0.314 (0.089)
Flowers with beetle frass (n /30) 1.733 (0.569) 1.567 (0.446) 2.400 (0.643) 3.233 (0.762)
Number of beetle larvae per plant (n /30) 0.733 (0.235) 0.667 (0.237) 1.300 (0.413) 1.267 (0.392)
Number of fruit per plant (n /32) 15.63 (1.344) 14.53 (1.638) 14.438 (1.279) 14.219 (1.536)
Number of failed fruit (n
/32) 6.938 (0.948) 7.281 (1.117) 9.125 (1.080) 8.250 (1.223)
Number of seeds per plant (n
/32) 1655.50 (186.08) 1752 (294.74) 1487.44 (221.91) 1445.00 (214.25)
Mean seed weight (mg) (n
/23) 0.149 (0.007) 0.139 (0.006) 0.128 (0.007) 0.129 (0.007)
506 OIKOS 110:3 (2005)