Ecology, 96(11), 2015, pp. 2902–2911� 2015 by the Ecological Society of America
Varying predator personalities generates contrasting preycommunities in an agroecosystem
RAPHAEL ROYAUTE1,3
AND JONATHAN N. PRUITT2
1Department of Biological Sciences, North Dakota State University, Fargo, North Dakota 58102 USA2Department of Biological Sciences, University of Pittsburgh, Pittsburgh, Pennsylvania 15260 USA
Abstract. Most taxa show consistent individual differences in behavior, a phenomenonoften referred to as animal ‘‘personalities.’’ While the links between individual personality andfitness have received considerable attention, little information is available on how animalpersonality impacts higher-order ecological processes, such as community dynamics. Using amesocosm experiment, we subjected a representative community of alfalfa pests to differentcompositions of personality types of the wolf spider Pardosa milvina. We show that subtlevariation in the personality composition of P. milvina populations generate wildly differentprey communities, where a mixture of both active and sedentary individuals performs best atsuppressing prey abundance. Our results provide the first experimental evidence that predatorpersonality types can generate contrasting prey communities. Moreover, our results suggestthat manipulating the representation of predator personality types may be a profitable avenueby which one can maximize the biocontrol potential of predator populations.
Key words: arthropod generalist predators; behavioral type 3 behavioral type interactions; biocontrol;Lycosidae; personality; pests community dynamics; predator–prey dynamics.
INTRODUCTION
Animal personality (also referred to as ‘‘tempera-
ment’’ or ‘‘behavioral syndromes’’) is a field of study
devoted to understanding the drivers and consequences
of consistent individual differences in behavior. Within
virtually any population, some individuals are consis-
tently more active, more aggressive, or are more willing
to engage in risk-taking behavior (Sih et al. 2004a, b,
Reale et al. 2007). Such personality differences are an
important determinant of an individual’s fitness (Dinge-
manse et al. 2004, Dingemanse and Reale 2005). These
individual differences have also been associated with
various underlying physiological and genetic mecha-
nisms (Careau et al. 2008, Bell and Aubin-Horth 2010,
Norton et al. 2011) and often exhibit a significant
heritability (Stirling et al. 2002). In addition, personality
differences have the potential to influence larger
ecological and evolutionary processes (Sih et al. 2012,
Wolf and Weissing 2012), including invasion success,
metapopulation dynamics, succession, species interac-
tions, and even extinction risk. In this manuscript, we
will refer to separate classes or categories of personalities
as behavioral types or personality types interchangeably.
The ecological consequences of animal personality
have been considered most rigorously in the context of
predator–prey interactions. Dozens of studies from a
wide variety of systems have demonstrated, for instance,
that prey personalities can influence prey growth rate
(reviewed in Stamps 2007), susceptibility to predation
(Bell and Sih 2007), and the intensity of trait-mediated
species interactions (Griffen et al. 2012, Toscano and
Griffen 2014). Other studies have shown that predator
personality types can influence predator feeding rate,
diet breadth, and the kinds of prey that individuals tend
to encounter and consume (Riechert and Hedrick 1993,
Matich et al. 2011; reviewed in Araujo et al. 2011). Still
other studies have shown that the outcome of predator–
prey interactions depend on subtle differences in the
representation of different personality types in both
predator and prey populations, hence the performance
of either predator or prey depends on the strategies
present in the opposing trophic level (Pruitt et al. 2012,
DiRienzo et al. 2013, McGhee et al. 2013, Sweeney et al.
2013). This last scenario is referred to as behavioral type
3 behavioral type interactions, and implies that no one
strategy in either predator or prey consistently experi-
ences superior performance. Yet, such interactions have
mostly been studied in the context of single-predator–
single-prey-species experiments and are therefore limited
in their ability to inform how predator personality
influences community dynamics.
A useful framework for predicting the community-
level consequences of behavioral type 3 behavioral type
interactions stems from the classic literature on foraging
mode. Huey and Pianka (1981) argued that the kinds of
prey consumed by predators would be different and
predictable depending on predators’ foraging mode:
sedentary predators should tend to catch active roving
prey whereas active predators would tend to capture
sedentary prey. Although the locomotor crossover
Manuscript received 17 December 2014; revised 2 April 2015;accepted 7 May 2015. Corresponding Editor: W. E. Snyder.
3 E-mail: [email protected]
2902
hypothesis has been mostly investigated in the context of
interspecific variation in foraging modes (Perry 2007, see
also Schmitz and Suttle 2001), a small number of studies
have revealed that similar dynamics can play out among
intraspecific trait variants in predator and prey (Pruitt et
al. 2012, DiRienzo et al. 2013, McGhee et al. 2013,
Sweeney et al. 2013) or in simulations (Scharf et al.
2006). Thus, an important question that remains to be
answered is whether variation in foraging mode within a
single predator species is of sufficient magnitude to
shape the interactions between a predator population
and myriad prey species, simultaneously. By extending
the locomotor crossover hypothesis to variation in
foraging mode within species, one might predict that
an assemblage of sedentary and active behavioral types
would be more effective at suppressing a community of
prey that exhibit a variety of locomotor and foraging
patterns. In contrast, we predict that if only a narrow
range of predators personality types are represented
within a population, then their top-down control on
prey assemblages should be weaker. In the study
described herein we probe more deeply into the role of
personality types in governing predator–prey interac-
tions to explore how the mixture of personality types in
populations of predators can impact entire communities
of prey using an agroecosystem model.
Spiders are ideal systems to explore the influence of
personality differences in ecological processes. Spiders
are dominant generalist arthropod predators in most
terrestrial ecosystems and can exert top-down control on
invertebrate prey assemblages, making them important
biocontrol agents in agroecosystems (Riechert and
Bishop 1990, Young and Edwards 1990, Carter and
Rypstra 1995). Their efficiency in biocontrol is also
mediated by a variety of behavioral processes, including
the timing of dispersal into agricultural habitats
(Hibbert and Buddle 2008, Sackett et al. 2009, Royaute
and Buddle 2012), intraguild predation and cannibalism
(Buddle 2002, Balfour et al. 2003, Buddle and Rypstra
2003), and the tendency to consume multiple prey items
beyond satiation (i.e., ‘‘wasteful killing’’; Maupin and
Riechert 2001). Fortuitously, spiders have also become a
model taxon for studying personality traits (Riechert
and Hedrick 1993, Johnson and Sih 2005, 2007, Pruitt et
al. 2008, Kralj-Fiser and Schneider 2012, Royaute et al.
2014) and personality differences are widespread across
more than a dozen families of spiders (reviewed in Pruitt
and Riechert 2012).
Our objective in this study was to test whether
personality differences in a generalist predator can
generate contrasting prey communities in an agro-
ecosystem. In particular, we focus on the extraordinarily
abundant wolf spider Pardosa milvina (see Plate 1). We
first classified individual P. milvina into either ‘‘active’’
or ‘‘inactive’’ personality types using a standardized
laboratory assay. We then deployed groups of active
spiders, inactive spiders, or a mixture of both active and
inactive spiders in experimental mesocosms that each
contained an identical assemblage of prey species. We
asked the following questions: (1) Does varying the
personality composition of the predator populations
generate contrasting prey communities? (2) If so, which
compositions (active, inactive, mixed) perform best at
diminishing which prey species or does one composition
work well universally? (3) Do some personality compo-
sitions generate more or less predictable prey commu-
nity structures? Addressing these questions is of broad
ecological interest for at least two reasons. First,
exploring the ecological consequences of personality
helps us understand when and how this aspect of
functional diversity trickles up to affect community
dynamics. Second, from a practical perspective, answer-
ing these questions can inform why integrative pest
management programs can succeed or fail, since
particular personality types (or combinations of person-
alities) may be more or less effective at controlling
particularly troublesome species of pests or entire pest
communities.
METHODS
Spider collection and laboratory maintenance.—Pardo-
sa milvina is a small wolf spider (Araneae: Lycosidae)
with high affinity for disturbed habitat and is extremely
common across many crop systems of North America
(Dondale and Redner 1990, Young and Edwards 1990,
Marshall and Rypstra 1999). P. milvina were collected as
late instar juveniles at an old field site in Powell,
Tennessee, USA in May–June 2010. Spiders were
collected at night using their eye shine. Spiders were
then transported back to laboratory at the University of
Tennessee, Knoxville. Spiders were housed individually
in 490-mL plastic cups containing a 2-cm pad of moist
sphagnum moss as substrate. Spiders were maintained
on a diet of size-matched domestic crickets provided ad
libitum twice weekly. Upon reaching maturity, we
selected only the females for use in our experiments,
ran them through a behavioral assay to assess their
activity level, measured their cephalothorax width at its
widest point using digital calipers as a proxy for body
size, and then assigned them to one of our experimental
treatments.
Activity level assay.—Open-field activity tests are
commonly used to test movement patterns in a
standardized manner (Carducci and Jakob 2000, Mon-
tiglio et al. 2010). These assays have been successfully
incorporated in personality studies to reveal patterns of
individual differences in activity and exploration strat-
egies in a wide variety of organisms, including spiders
(Dochtermann and Nelson 2014, Montiglio et al. 2014,
Royaute et al. 2014). Activity assays were conducted by
placing a single P. milvina in a clean 18 3 18 cm plastic
arena lined with 1 3 1 cm graphing paper. We then
covered the spider with a black cover object (4 cm high,
8 cm radius). After 30 s acclimation, we removed the
cover object and placed the lid on the arena. We waited
until the spider started movement and then counted the
November 2015 2903PREDATOR PERSONALITY AND PREY COMMUNITY
number of 1-cm lines crossed over the next 3 minutes.
We deemed individuals that traversed further in these
environments as more active. To ensure that this
behavior was a repeatable trait of individual spiders,
we measured a pool of 19 P. milvina once daily for eight
consecutive days. This pool of individuals was not used
in our mesocosm studies.
Single-predator–single-prey assays.—A pool (n ¼ 78)
of mature female P. milvina were collected to test
whether differences in predators’ willingness to accept
various prey species was associated with their personal-
ity type. Females were starved two weeks before being
presented with a single prey item in a structurally
depauperate arena. These arenas consisted of a 490-mL
plastic deli container lined with filter paper. The arena
perimeter was lined with petroleum jelly 2 cm above the
base, in order to prevent prey items from escaping into
the tops of the containers. A single spider of known
activity level was first released into the arena and given
20 minutes to acclimate. A single prey item was then
placed centrally within the arena using forceps. We left
the spider to interact with the prey for the next 24 hours.
At the end of this time we recorded whether the spider
consumed the prey item or not. Instances where the prey
item escaped into the top of the container we removed
from our analysis. A separate group of spiders was used
for each prey species. Sample sizes for each prey species
were as follows: blister beetles (n ¼ 12), potato
leafhoppers (n ¼ 12), beet armyworms (n ¼ 12), pea
aphids (n ¼ 14), sharpshooters (n ¼ 14), and alfalfa
weevils (n ¼ 14).
Alfalfa establishment.—We selected a 5 3 20 m plot
with direct sun atop a hill to seed our alfalfa. Prior to
planting, we cleaned the area, worked the soil, and
removed any and all debris. Certified organic alfalfa
seeds (Sustainable Seed Company, Covelo, California,
USA) were planted by sprinkling seed atop the bare soil
and covering the seeds with an additional 2 cm of loose
soil. Seeds were planted in rows spaced 32 cm apart,
with approximately 115 g of seeds deployed every 7 m.
Seeds were permitted to sprout and grow until they
reached approximately 18 cm in height. Throughout this
initial growth phase, we meticulously checked the alfalfa
beds for weeds and insects twice daily, since young
alfalfa beds are extremely susceptible to infestation by
potato leafhoppers (Empoasca fabae).
Pest species collection and release.—Pest species were
collected opportunistically on nearby alfalfa fields,
roadsides, and old field habitats by sweep netting. We
focused on six of alfalfa’s most voracious insect pests:
blister beetles (Epicauta sp.), potato leafhoppers (Em-
poasca fabae), pea aphids (Acyrthosiphon pisum), sharp-
shooters (Xyphon), alfalfa weevils (Hypera postica), and
beet armyworms (Spodoptera exigua). Moreover, each
of these pest species possesses distinct activity patterns,
varying from sap-suckers with relatively sedentary
foraging mode (pea aphids, potato leafhoppers, and
sharpshooters) to more free-living leaf chewers that are
generally more active (armyworms, alfalfa weevils, and
blister beetles [Price et al. 2011]). These biological
differences allowed to test whether specific predator
personality types are more likely to consume certain
prey species. Stock populations of pests were housed in
52 3 52 3 52 cm Lumite mesh mesocosms (BioQuip
1450DSV; Rancho Dominguez, California, USA),
which were placed over beds of alfalfa. We were unable
to collect larval S. exigua of a suitable size for P. milvina
to devour. Thus, we purchased our S. exigua (instar 2–3)
from a nearby hobbyist herpetologist.
Mesocosm experiment.—Fifty-five 52 3 52 3 52 cm
Lumite mesh mesocosms (BioQuip 1450DSV) were
deployed shortly after the alfalfa reached approximately
18 cm in height. Mesocosms were deployed by unscrew-
ing and removing their bottom panels and setting them
atop two adjacent rows of alfalfa. We then trimmed and
adjusted the alfalfa that was immediately adjacent to the
sides so that the mesocosms set flat against the ground.
We then placed an 18 cm skirt of plastic sheeting around
the edge of the mesocosm and molded a layer of potting
soil around the base to seal any remaining gaps between
the bottom of the mesocosm and the ground. Meso-
cosms were stocked with a standardized community of
pest arthropods and P. milvina. The pest community
consisted of 15 blister beetles, 15 potato leafhoppers, 10
beet armyworms, 9 pea aphids, 5 sharpshooters, and 8
alfalfa weevils. These numbers were selected based on
the number of prey caught using sweep nets in adjacent,
untreated alfalfa beds and old field habitats. Predator
populations consisted of eight active, eight inactive, or a
mixture of four inactive and four active P. milvina.
Spiders that exhibited activity scores in the top 50% of
the distribution of activity levels were labeled ‘‘active,’’
and spiders that exhibited activity scores in the bottom
50% of the distribution were labeled ‘‘inactive.’’ This
classification resulted in significant differences in activity
levels among all treatments (Appendix: Fig. A1). We
elected not to create predator populations with more
extreme compositions (e.g., only the top and bottom
deciles) because we were interested in creating distribu-
tions of predator personality types that might commonly
be found together in nature. Predators were starved for
two weeks prior to being deployed within the meso-
cosms. We deployed 15 mesocosms of each treatment
(active, inactive, mixture) plus 10 no-spider controls
(total n¼ 55). The latches to the tops of the mesocosms
were then closed, and we allowed the experimental
community to interact for seven days. At the end of this
time, we collected all of the surviving pests and noted the
resulting community structures and the number of P.
milvina that were recovered at the end of the experiment.
Statistical analyses.—All analyses were performed in
R version 3.0.2 (R Development Core Team 2013).
We first determined the repeatability of activity using
a linear mixed model with the package lme4 (Bates et al.
2014). We included body size and measurement number
as a fixed effect, individuals as random effects, and
RAPHAEL ROYAUTE AND JONATHAN N. PRUITT2904 Ecology, Vol. 96, No. 11
scaled activity values to standard deviation units to ease
convergence. The repeatability of activity was calculated
from the random component of the mixed model as R¼VID/(VID þ VR), where VID represents the among-
individual variance and VR represents the within-
individual, or residual variance. The 95% confidence
intervals (CI) were calculated using bootstrapping with
1000 simulation to assess the magnitude and precision of
activity repeatability and the significance of fixed effects.
We used generalized linear models (GLM) with
Poisson error to evaluate the effects of spider predation
from treatment groups with differing personality com-
positions on pest abundance. Each of the six pest species
was included in a separate GLM with pest abundance
used as the dependent variable and treatment group
included as a fixed effect. We used likelihood ratio tests
to test for an overall treatment effect and Tukey’s
multiple comparison test to detect significant differences
among treatments. We further compared the overlap of
95% CI in the mixture treatment with the average
predation in the active and inactive treatments combined
in order to test for the presence of risk enhancement
(lower survival than that expected by the combined
effect of both phenotypes), risk reduction (higher
survival than that expected by the combined effect of
both phenotypes) or substitutive effects (lack of risk
enhancement or reduction) in the mixture treatment (Sih
et al. 1998). We used a similar procedure for determining
potential differences in cannibalism among treatments.
In order to test for an effect of predator behavioral
type and prey species on capture propensity in single
predator–prey assay, we used a GLM with binomial
error structure. As above, we used likelihood ratio tests
to estimate the strength of the spider activity 3 prey
species interaction. We used nonmetric multidimension-
al scaling (NMDS) in the package vegan (Oksanen et al.
2007) to compare pest community composition between
treatment groups. NMDS is a nonparametric ordination
technique that allows robust representation of commu-
nity structures while not requiring linear relationships
between variables (McCune et al. 2002). Abundance
data were log10(x þ 1)-transformed to reduce the
influence of dominant species and results are based on
the Bray-Curtis dissimilarity matrix. To further test the
effects of treatment group on pest community compo-
sition, we used PERMANOVA to detect an effect of
treatment on average community composition (i.e., to
compare whether treatments differed in their centroids)
and tested for multivariate homogeneity of variance in
community composition between treatments (i.e., to
compare whether treatments differed in their distance to
their centroids).
RESULTS
Activity had moderate repeatability (R ¼ 0.36 [0.30,
0.44] [95% CI]) with similar values to that reported in the
behavioral literature for this trait (Bell et al. 2009). Note
that neither individuals’ body size nor measurement
number was associated with their activity levels (body
size¼�0.003 [�1.44, 1.27]; measurement number¼ 0.03
[�0.02, 0.09]). Taken together, these results demonstrate
that individuals showed consistent differences in activity
levels that are independent of body size and the number
of behavioral measurements (Appendix: Fig. A2).
Spiders’ tendency to attack the various prey species in
isolation was not significantly related to spider activity
(LRT, v2 ¼ 1.358, P ¼ 0.24), nor did we found any
indication of a significant species 3 spider activity
interaction (LRT, v2 ¼ 7.307, P ¼ 0.20). We did,
however, detect strong indication of difference in prey
preference (LRT, v2 ¼ 32.366, P , 0.0001). Potato
leafhoppers and sharpshooters were both more strongly
preyed upon (capture probability .0.5), while alfalfa
weevils, beet armyworms, and pea aphids were generally
rejected in single-predator–single-prey encounters (cap-
ture probability , 0.3; Fig. 1).
Based on the number of spiders recovered at the end
of the experiment, we did not detect any significant
difference in cannibalism rates among treatments (P ¼0.72; Fig. A3). With the exception of beet armyworms,
all pests showed a substantial decrease in abundance
when spiders were added (P , 0.05; Table 1). Three
species (alfalfa weevils, blister beetles, and sharpshoot-
ers) showed significantly lower pest abundance with a
mix of active and inactive personality types (Fig. 2).
These species also showed evidence of risk enhancement
effects since the prey survival in the mixed treatment was
lower than that expected by the sum of the two
personality types (mean pest survival [CI]; alfalfa
weevils, mixed treatment, 3.27 [2.47, 4.32], expected,
5.87; blister beetles, mixed treatment, 6.13 [5.00, 7.52],
expected, 10; sharpshooters, mixed treatment, 2.20 [1.56,
FIG. 1. Effect of predator behavioral type on capturepropensity for each prey species. Values represent averageprobability of capture with 95% confidence intervals based onpredicted values from generalized linear models. Pest species areAw, alfalfa weevil; Ba, beet armyworm; Bb, blister beetle; Pa,pea aphids; Pl, potato leafhopper; and Sh, sharpshooter.
November 2015 2905PREDATOR PERSONALITY AND PREY COMMUNITY
3.09], expected, 3.50). Potato leafhoppers showed no
differences in abundance between mixed and active
groups but both differed significantly from the inactive
treatment. This suggests that either the active pheno-
types contributed most to the control of this species or
that interference among personality types reduced pest
suppression in the mixed treatment. Pea aphids had
significantly lower abundance in all spider treatments,
suggesting that all personality types contributed equally
to the control of this species. In addition, personality
composition generated differences in pest abundance
community composition with the mixed group being the
most distinct from all other treatments (Fig. 3). This
result was corroborated by a significant treatment effect
in the PERMANOVA (P , 0.001). However, the mixed
treatment also showed highest variance in community
composition when performing the multivariate homo-
geneity of variance test (Table 2, P , 0.0001). This
means that the mixed phenotype produced the most
variable results in terms of reduction of pest abundance.
DISCUSSION
The objective of our experiment was to understand
whether varying the representation of predator person-
ality types has an effect on prey community structure.
Our results show that prey regulation depended highly
on the personality types of predators: varying the
representation of active vs. sedentary P. milvina
generated wildly different prey communities over the
course of only a few days. For one-half of the pest
species that we tested, deploying a mixture of active and
inactive personality types led to significant risk enhance-
ment effects compared to monotypic predator groups.
Single-predator–single-prey encounters revealed that
different personality types are equally willing to attack
prey when they are presented to them, thus suggesting
TABLE 1. Effects of phenotypic group composition on pest abundance post treatment.
Speciesand term Estimate SE Z P
LRT
ComparisonsStatistic P
Blister beetles (Epicauta sp.)
Intercept 2.54 0.09 28.64 ,0.0001 31.04 ,0.0001 inactive , controlInactive �0.25 0.12 �2.09 ,0.05 mixed , control***Mixed �0.73 0.14 �5.32 ,0.0001 active , controlActive �0.23 0.12 �1.88 0.06 mixed , inactive**, active . inactive,
active . mixed***
Potato leafhoppers (Empoasca fabae)
Intercept 2.46 0.09 26.61 ,0.0001 64.76 ,0.0001 inactive , controlInactive �0.31 0.13 �2.41 ,0.05 mixed , control***Mixed �0.85 0.15 �5.75 ,0.0001 active , control***Active �1.09 0.16 �6.83 ,0.0001 mixed , inactive***, active ,
inactive***, active , mixed
Pea aphids (Aphis pisum)
Intercept 4.23 0.04 110.63 ,0.0001 44.75 ,0.0001 inactive , control***Inactive �0.34 0.05 �6.47 ,0.0001 mixed , control***Mixed �0.27 0.05 �5.13 ,0.0001 active , control***Active �0.24 0.05 �4.65 ,0.0001 mixed , inactive, active . inactive,
active . mixed
Sharpshooters (Xyphon sp.)
Intercept 1.46 0.15 9.57 ,0.0001 9.6 ,0.05 inactive , controlInactive �0.25 0.21 �1.22 0.22 mixed , control*Mixed �0.67 0.23 �2.9 ,0.005 active , controlActive �0.16 0.2 �0.78 0.43 mixed , inactive, active . inactive,
active . mixed
Alfalfa weevils (Hypera postica)
Intercept 1.81 0.13 14.12 ,0.0001 16.74 ,0.001 inactive , controlInactive 0.01 0.17 0.03 0.97 mixed , control**Mixed �0.62 0.19 �3.26 ,0.005 active , controlActive �0.09 0.17 �0.51 0.61 mixed , inactive**, active , inactive,
active . mixed*
Beet armyworms (Spodoptera exigua)
Intercept 1.99 0.12 16.98 ,0.0001 1.33 0.72 inactive . controlInactive 0.07 0.15 0.5 0.62 mixed , controlMixed �0.04 0.15 �0.28 0.78 active , controlActive �0.07 0.15 �0.46 0.64 mixed , inactive, active , inactive,
active , mixed
Notes: Values represent results of generalized linear models with Poisson error with the control group set as the reference (i.e.,the intercept). Boldface type indicates significance for a , 0.05. Abbreviations are SE, standard error associated with the estimate,and LRT, likelihood ratio test value for the treatment term. Comparisons based on Tukey HSD test.
* P , 0.05; ** P , 0.01; *** P , 0.001.
RAPHAEL ROYAUTE AND JONATHAN N. PRUITT2906 Ecology, Vol. 96, No. 11
that the patterns observed in our mesocosms reflect
more than variation in individuals’ willingness to attack
prey or personality-dependent differences in diet
breadth. When investigating the effects of predator
personality composition at the community level, our
results reveal that deploying a group of only active
predators or combination of both active and inactive
predators created communities with no overlap with our
no-predator controls (Fig. 2). Yet, we also noted that
the active/inactive mixture treatment generated the
greatest variety of experimental outcomes (i.e., the
outcomes were less predictable and more broadly
scattered), whereas the inactive-only and active-only
treatments had very consistent outcomes across repli-
cates.
The fact that a mixture of active and inactive predator
types performs best at suppressing a variety of prey
populations is notable. Although we failed to observe
any instance where sedentary P. milvina were more
effective than active individuals in suppressing any one
prey species, a mixture of active and inactive individuals
was always at least as effective as monotypic groups at
suppressing prey populations and often performed much
better. Similar consequences of intraspecific behavioral
FIG. 2. Changes in pest abundance in relation to phenotypic group composition. Values represent average pest abundance post-treatment with 95% confidence intervals based on predicted values from generalized linear models. Solid line represents pestabundance at the beginning of the experiment, dotted line represents the expected performance of the mixed treatment (calculatedas the average of the active and inactive treatments). Treatments are C, control; I, inactive; M, mixed; and A, active.
FIG. 3. Change in pest community composition dependingon spider phenotypic group. Values are based on nonmetricmultidimensional scaling (NMDS) ordination with 95% densityellipses overlain for each group. Stress¼ 0.15.
November 2015 2907PREDATOR PERSONALITY AND PREY COMMUNITY
variation in predators have been noted in at least one
other experimental system (Finke and Snyder 2008).
Yet, a closer examination of the pest species most
affected by sedentary vs. active foraging types reveals
more curious patterns, which appear at odds with Huey
and Pianka’s predictions (1981). Sap-sucker species tend
to spend extended periods of time with their stylet
inserted into plant vascular tissues (Auclair 1963,
Weintraub and Beanland 2006), yet none of these
species showed evidence for increased predation pres-
sure in the active predator treatment. Similarly, free-
roaming leaf chewers either showed evidence of risk
enhancement in the mixed predator treatment (alfalfa
weevils, blister beetles) or no significant effect of spider
predation (beet armyworms). Overall, it seems that the
results of results of variation in predator personality
composition yields more predictive power at the
community level rather than by examining the effects
on individual prey species.
These details are important to consider because the
outcome of predatory encounters can depend on a
multitude of factors, including prey defenses (e.g., alarm
pheromones and dropping behavior in aphids; Nault et
al. 1973, Losey and Denno 1998) or changes in prey
habitat use in response to predator presence (Schmitz
2005). Current theory on predator–prey dynamics
predicts that the direction and magnitude of multiple
predator effects (MPE; risk enhancement/reduction or
substitutive effects) will depend strongly on the interac-
tion between predator and prey foraging mode as well as
the overlap in habitat domain among predator and prey
species (Schmitz 2007). Risk enhancement is expected to
happen whenever prey species have narrow habitat
domains and their predators have broad and overlap-
TABLE 2. Effect of predator personality type on pest community composition based onPERMANOVA.
Response df F P R2 Comparisons
Treatment 3 9.05 ,0.001 0.35 inactive . controlResiduals 51 mixed . control****, active . control,
mixed . inactive***, active . inactive,active , mixed***
Note: Comparisons based on Tukey HSD test indicating multivariate homogeneity of groupdispersion among treatments.
**** P , 0.0001; *** P , 0.001.
PLATE 1. Male Pardosa milvina. Photo credit: Matthew Persons.
RAPHAEL ROYAUTE AND JONATHAN N. PRUITT2908 Ecology, Vol. 96, No. 11
ping habitat domains and similar hunting strategies. In
this configuration, prey cannot find predator-free space
and predators encounter little risk of intraguild preda-
tion or interference as they do not need to rely
exclusively on a unique prey species (Griffen 2006). P.
milvina has a relatively broad habitat domain (Sitvarin
and Rypstra 2014) but how individual variation in
activity level relates to habitat use remains unknown in
this species. We do not have strong data on the habitat
domains of our various prey species either. Yet, we
detected risk enhancement for three prey species (alfalfa
weevils, blister beetles, and sharpshooters) in our mixed
treatment, thus theory predicts that active and sedentary
P. milvina should have broad and overlapping habitat
domains and that these three prey species will exhibit
narrow habitat domains (Schmitz 2007). Taken togeth-
er, given the necessary data, we reason that the MPE
framework could become a powerful tool for predicting
the relative influence of intra- and inter-specific variation
in predator foraging mode in shaping prey communities.
Some mixtures of predator personality types generat-
ed more variable outcomes than others. While both the
active-only and inactive-only treatments tended to have
very similar outcomes on prey community structure
across replicates, the mixed treatment created wildly
different communities across replicates. This is an
intriguing finding because it suggests that intraspecific
variation in predator populations could be a major
determinant of spatiotemporal variation in prey com-
munity structure, where the same mix of predator
personality types could generate surprisingly different
prey communities across space and time. In contrast,
given our results, one would predict that the erosion of
intraspecific variation in apex predators would generate
a homogenization of prey community composition
across space and time, even when predator abundance
remains constant. This, in turn, has important conser-
vation implications and casts doubt on conventional
management schemes that tend to focus on species
abundance and/or persistence and tend to discount
subtle shifts in key functional traits, like behavior
(Watters et al. 2003, Watters and Meehan 2007).
The findings herein have intriguing ecological impli-
cations because they suggest that any factor that
imposes a selective pressure on predator personality
types (e.g., human-induced rapid environment change,
selective hunting, climate change) is likely to have non-
intuitive consequences for lower trophic levels. For
instance, any selective pressure that reduces trait
variation in predator populations (e.g., via hunting
and poaching, Biro and Post 2008) could yield surprising
rises in the abundance of various prey species, even if
predator population numbers remain largely consistent
through time, since an assemblage of personality types is
predicted to suppress prey communities more effectively
than an equal number of monotypic individuals (Fig. 1).
Likewise, directional selection on predator personality
types could result in either increases or decreases in the
abundance of various prey species, depending on the
direction of selection considered.
One of the most important consequences of our study
is that it brings the animal personality framework to
bear on a classic question in applied ecology: When and
how can natural enemies be used to combat pests? In a
generic sense, our results demonstrate that varying the
personality type of predators has the potential to change
their efficacy as biocontrol agents. Consistent with the
established theory on animal personalities, we show that
a mix of personality types best suppresses an assemblage
of pest species, although the pest communities generated
were also highly variable. Our results lead to several
interesting recommendations for integrative pest man-
agement (IPM) plans. For instance, IPM plans may
benefit by promoting selection regimes that preserve
intraspecific variation in their key and ancillary control
agents. Yet, an alternative could be to deploy manage-
ment practices that favor the persistence of active
individuals, since in our system the active-only treatment
diminished pest numbers reasonably well and generated
highly stereotyped pest communities. Consequently, in
terms of biocontrol potential, it seems the ‘‘optimal’’
personality composition of P. milvina populations
depends on the precise goal in mind: diminish a
particular pest species to the lowest level possible (or
some ‘‘ideal’’ level), diminish the broadest suite of pests
to the lowest levels possible (or some ‘‘ideal’’ level), or
generate a highly predictable prey community, which
may be easier to control with additional regulatory
mechanisms.
It is, of course, important to note that top-down
control on pests resulting from spiders and other
generalist predators is a community process provided
by all species present in an agroecosystem, rather than a
single species (Riechert and Bishop 1990, Symondson et
al. 2002). Intra-guild predation and cannibalism are
common phenomena in such systems and also reduce the
top-down effect of predator assemblages (Buddle 2002,
Balfour et al. 2003). Thus, we argue that an important
way forward would be to quantify which types of
personality compositions have the largest impact on pest
communities whilst minimizing undesired effects (e.g.,
competition within guilds) by considering personality
types in multiple predator species simultaneously.
CONCLUSIONS
We show that altering the composition of personality
types in a top arthropod predator can magnify their top-
down effects by 200–300% (Fig. 1). A mixture of active
and sedentary spiders provided the best pest regulation
service for three of the six pest species tested and the
resulting pest community structure was most variable
with this treatment (Fig. 2). Understanding how this
variation among replicates comes about is essential in
order to move this field forward. These findings also
have practical importance as they show a clear link
November 2015 2909PREDATOR PERSONALITY AND PREY COMMUNITY
between personality composition and the performance
of natural enemies in pest biocontrol.
ACKNOWLEDGMENTS
We thank Jamie Troupe, Austin Dillon, and Jeffrey Taylorfor assistance collecting wolf spiders and the establishment ofour mesocosms. We are indebted to the University ofPittsburgh’s Department of Biological Sciences and theUniversity of Tennessee’s Department of Ecology and Evolu-tionary Biology for support for this research. We thank Ned A.Dochtermann for comments on an earlier draft of this article.We also thank William Snyder and two anonymous reviewersfor their thoughtful comments.
LITERATURE CITED
Araujo, M. S., D. I. Bolnick, and C. A. Layman. 2011. Theecological causes of individual specialisation. Ecology Letters14:948–958.
Auclair, J. L. 1963. Aphid feeding and nutrition. AnnualReview of Entomology 8:439–490.
Balfour, R. A., C. M. Buddle, A. L. Rypstra, S. E. Walker, andS. D. Marshall. 2003. Ontogenetic shifts in competitiveinteractions and intra-guild predation between two wolfspider species. Ecological Entomology 28:25–30.
Bates, D., M. Maechler, B. Bolker, and S. Walker. 2014. lme4:linear mixed-effects models using Eigen and S4. R packageversion 1.1-6. http://cran.r-project.org/web/packages/lme4/index.html
Bell, A. M., and N. Aubin-Horth. 2010. What can wholegenome expression data tell us about the ecology andevolution of personality? Philosophical Transactions of theRoyal Society B 365:4001–4012.
Bell, A. M., S. J. Hankison, and K. L. Laskowski. 2009. Therepeatability of behaviour: a meta-analysis. Animal Behav-iour 77:771–783.
Bell, A. M., and A. Sih. 2007. Exposure to predation generatespersonality in threespined sticklebacks (Gasterosteus aculea-tus). Ecology Letters 10:828–834.
Biro, P. A., and J. R. Post. 2008. Rapid depletion of genotypeswith fast growth and bold personality traits from harvestedfish populations. Proceedings of the National Academy ofSciences USA 105:2919–2922.
Buddle, C. M. 2002. Interactions among young stages of thewolf spiders Pardosa moesta and P. mackenziana (Araneae:Lycosidae). Oikos 96:130–136.
Buddle, C. M., and A. L. Rypstra. 2003. Factors initiatingemigration of two wolf spider species (Araneae: Lycosidae) inan agroecosystem. Environmental Entomology 32:88–95.
Carducci, J. P., and E. M. Jakob. 2000. Rearing environmentaffects behaviour of jumping spiders. Animal Behaviour 59:39–46.
Careau, V., D. Thomas, M. M. Humphries, and D. Reale. 2008.Energy metabolism and animal personality. Oikos 117:641–653.
Carter, P. E., and A. L. Rypstra. 1995. Top-down effects insoybean agroecosystems: spider density affects herbivoredamage. Oikos 72:433–439.
Dingemanse, N. J., C. Both, P. J. Drent, and J. M. Tinbergen.2004. Fitness consequences of avian personalities in afluctuating environment. Proceedings of the Royal SocietyB 271:847–852.
Dingemanse, N. J., and D. Reale. 2005. Natural selection andanimal personality. Behaviour 142:1159–1184.
DiRienzo, N., J. N. Pruitt, and A. V. Hedrick. 2013. Thecombined behavioural tendencies of predator and preymediate the outcome of their interaction. Animal Behaviour86:317–322.
Dochtermann, N. A., and A. B. Nelson. 2014. Multiple facetsof exploratory behavior in house crickets (Acheta domes-
ticus): split personalities or simply different behaviors?Ethology 120:1110–1117.
Dondale, C. D., and J. H. Redner. 1990. The insects andarachnids of Canada. Part 17. The wolf spiders, nurserywebspiders, and lynx spiders of Canada and Alaska. Araneae:Lycosidae, Pisauridae, and Oxyopidae. Agriculture CanadaPublications, Ottawa, Ontario, Canada.
Finke, D. L., and W. E. Snyder. 2008. Niche partitioningincreases resource exploitation by diverse communities.Science 321:1488–1490.
Griffen, B. D. 2006. Detecting emergent effects of multiplepredator species. Oecologia 148:702–709.
Griffen, B. D., B. J. Toscano, and J. Gatto. 2012. The role ofindividual behavior type in mediating indirect interactions.Ecology 93:1935–1943.
Hibbert, A. C., and C. M. Buddle. 2008. Assessing the dispersalof spiders within agricultural fields and an adjacent matureforest. Journal of Arachnology 36:195–198.
Huey, R. B., and E. R. Pianka. 1981. Ecological consequencesof foraging mode. Ecology 62:991–999.
Johnson, J. C., and A. Sih. 2005. Precopulatory sexualcannibalism in fishing spiders (Dolomedes triton): a role forbehavioral syndromes. Behavioral Ecology and Sociobiology58:390–396.
Johnson, J. C., and A. Sih. 2007. Fear, food, sex and parentalcare: a syndrome of boldness in the fishing spider, Dolomedestriton. Animal Behaviour 74:1131–1138.
Kralj-Fiser, S., and J. M. Schneider. 2012. Individual behav-ioural consistency and plasticity in an urban spider. AnimalBehaviour 84:197–204.
Losey, J. E., and R. F. Denno. 1998. Positive predator–predator interactions: enhanced predation rates and syner-gistic suppression of aphid populations. Ecology 79:2143–2152.
Marshall, S. D., and A. L. Rypstra. 1999. Spider competition instructurally simple ecosystems. Journal of Arachnology 343–350.
Matich, P., M. R. Heithaus, and C. A. Layman. 2011.Contrasting patterns of individual specialization and trophiccoupling in two marine apex predators. Journal of AnimalEcology 80:294–305.
Maupin, J. L., and S. E. Riechert. 2001. Superfluous killing inspiders: a consequence of adaptation to food-limitedenvironments? Behavioral Ecology 12:569–576.
McCune, B., J. B. Grace, and D. L. Urban. 2002. Analysis ofecological communities. MjM Software Design, GlenedenBeach, Oregon, USA.
McGhee, K. E., L. M. Pintor, and A. M. Bell. 2013. Reciprocalbehavioral plasticity and behavioral types during predator–prey interactions. American Naturalist 182:704–717.
Montiglio, P.-O., D. Garant, P. Bergeron, G. Dubuc Messier,and D. Reale. 2014. Pulsed resources and the couplingbetween life-history strategies and exploration patterns ineastern chipmunks (Tamias striatus). Journal of AnimalEcology 83:720–728.
Montiglio, P.-O., D. Garant, D. Thomas, and D. Reale. 2010.Individual variation in temporal activity patterns in open-field tests. Animal Behaviour 80:905–912.
Nault, L. R., L. J. Edwards, and W. E. Styer. 1973. Aphidalarm pheromones: secretion and reception. EnvironmentalEcology 2:101–105.
Norton, W. H. J., K. Stumpenhorst, T. Faus-Kessler, A.Folchert, N. Rohner, M. P. Harris, J. Callebert, and L. Bally-Cuif. 2011. Modulation of Fgfr1a signaling in zebrafishreveals a genetic basis for the aggression–boldness syndrome.Journal of Neuroscience 31:13796–13807.
Oksanen, J., R. Kindt, P. Legendre, B. O’Hara, M. H. H.Stevens, M. J. Oksanen, and M. Suggests. 2007. The veganpackage. Community ecology package. http://cran.r-project.org/web/packages/vegan/index.html
RAPHAEL ROYAUTE AND JONATHAN N. PRUITT2910 Ecology, Vol. 96, No. 11
Perry, G. 2007. Movement patterns in lizards: measurement,modality and behavioral correlates. Pages 13–48 in S. M.Reilly, L. D. McBrayer, and D. B. Miles, editors. Lizardecology. Cambridge University Press, Cambridge, UK.
Price, P. W., R. F. Denno, M. D. Eubanks, D. L. Finke, and I.Kaplan. 2011. Insect ecology: behavior, populations andcommunities. Cambridge University Press, Cambridge, UK.
Pruitt, J. N., and S. E. Riechert. 2012. The ecologicalconsequences of temperament in spiders. Current Zoology58:589–596.
Pruitt, J. N., S. E. Riechert, and T. C. Jones. 2008. Behaviouralsyndromes and their fitness consequences in a sociallypolymorphic spider, Anelosimus studiosus. Animal Behaviour76:871–879.
Pruitt, J. N., J. J. Stachowicz, and A. Sih. 2012. Behavioraltypes of predator and prey jointly determine prey survival:potential implications for the maintenance of within-speciesbehavioral variation. American Naturalist 179:217–227.
R Development Core Team. 2013. R: a language andenvironment for statistical computing. R Project for Statis-tical Computing, Vienna, Austria. www.r-project.org
Reale, D., S. M. Reader, D. Sol, P. T. McDougall, and N. J.Dingemanse. 2007. Integrating animal temperament withinecology and evolution. Biological Reviews 82:291–318.
Riechert, S. E., and L. Bishop. 1990. Prey control by anassemblage of generalist predators: spiders in garden testsystems. Ecology 71:1441–1450.
Riechert, S. E., and A. V. Hedrick. 1993. A test for correlationsamong fitness-linked behavioural traits in the spider Agele-nopsis aperta (Araneae, Agelenidae). Animal Behaviour 46:669–675.
Royaute, R., and C. M. Buddle. 2012. Colonization dynamicsof agroecosystem spider assemblages after snow-melt inQuebec (Canada). Journal of Arachnology 40:48–58.
Royaute, R., C. M. Buddle, and C. Vincent. 2014. Interpop-ulation variations in behavioral syndromes of a jumpingspider from insecticide-treated and insecticide-free orchards.Ethology 120:127–139.
Sackett, T. E., C. M. Buddle, and C. Vincent. 2009. Dynamicsof spider colonization of apple orchards from adjacentdeciduous forest. Agriculture, Ecosystems and Environment129:144–148.
Scharf, I., E. Nulman, O. Ovadia, and A. Bouskila. 2006.Efficiency evaluation of two competing foraging modes underdifferent conditions. American Naturalist 168:350–357.
Schmitz, O. J. 2005. Behavior of predators and prey and linkswith population level processes. Pages 256–278 in P. Barbosaand I. Castellanos, editors. Ecology of predator–preyinteractions. Oxford University Press, Oxford, UK.
Schmitz, O. J. 2007. Predator diversity and trophic interactions.Ecology 88:2415–2426.
Schmitz, O. J., and K. B. Suttle. 2001. Effects of top predatorspecies on direct and indirect interactions in a food web.Ecology 82:2072–2081.
Sih, A., A. Bell, and J. C. Johnson. 2004a. Behavioralsyndromes: an ecological and evolutionary overview. Trendsin Ecology & Evolution 19:372–378.
Sih, A., A. M. Bell, J. C. Johnson, and R. E. Ziemba. 2004b.Behavioral syndromes: an integrative overview. QuarterlyReview of Biology 79:241–277.
Sih, A., J. Cote, M. Evans, S. Fogarty, and J. Pruitt. 2012.Ecological implications of behavioural syndromes. EcologyLetters 15:278–289.
Sih, A., G. Englund, and D. Wooster. 1998. Emergent impactsof multiple predators on prey. Trends in Ecology &Evolution 13:350–355.
Sitvarin, Michael I., and A. L. Rypstra. 2014. The importanceof intraguild predation in predicting emergent multiplepredator effects. Ecology 95:2936–2945.
Stamps, J. A. 2007. Growth–mortality tradeoffs and ‘‘person-ality traits’’ in animals. Ecology Letters 10:355–363.
Stirling, D. G., D. Reale, and D. A. Roff. 2002. Selection,structure and the heritability of behaviour. Journal ofEvolutionary Biology 15:277–289.
Sweeney, K., B. Cusack, F. Armagost, T. O’Brien, C. N.Keiser, and J. N. Pruitt. 2013. Predator and prey activitylevels jointly influence the outcome of long-term foragingbouts. Behavioral Ecology 24:1205–1210.
Symondson, W. O., K. D. Sunderland, and M. H. Greenstone.2002. Can generalist predators be effective biocontrol agents?Annual Review of Entomology 47:561–594.
Toscano, B. J., and B. D. Griffen. 2014. Trait-mediatedfunctional responses: predator behavioural type mediatesprey consumption. Journal of Animal Ecology 83:1469–1477.
Watters, J. V., S. C. Lema, and G. A. Nevitt. 2003. Phenotypemanagement: a new approach to habitat restoration.Biological Conservation 112:435–445.
Watters, J. V., and C. L. Meehan. 2007. Different strokes: Canmanaging behavioral types increase post-release success?Applied Animal Behaviour Science 102:364–379.
Weintraub, P. G., and L. Beanland. 2006. Insect vectors ofphytoplasmas. Annual Review of Entomology 51:91–111.
Wolf, M., and F. J. Weissing. 2012. Animal personalities:consequences for ecology and evolution. Trends in Ecology &Evolution 27:452–461.
Young, O., and G. Edwards. 1990. Spiders in United Statesfield crops and their potential effect on crop pests. Journal ofArachnology 18:1–27.
SUPPLEMENTAL MATERIAL
Ecological Archives
The Appendix is available online: http://dx.doi.org/10.1890/14-2424.1.sm
November 2015 2911PREDATOR PERSONALITY AND PREY COMMUNITY
Top Related