MIAMI UNIVERSITY
The Graduate School
Certificate for Approving the Dissertation
We hereby approve the Dissertation
of
Michael Sitvarin
Candidate for the Degree:
Doctor of Philosophy
_________________________
Director
Ann Rypstra
_________________________
Reader
Tom Crist
_________________________
Reader
Brian Keane
_________________________
Reader
Nancy Solomon
_________________________
Graduate School Representative
Dave Gorchov
ABSTRACT
BEHAVIORAL AND ECOLOGICAL CONSEQUENCES OF MULTIPLE INTRAGUILD
PREDATORS AND CONNECTIONS BETWEEN PREDATORS, PREY, AND ECOSYSTEM
FUNCTION
by Michael I. Sitvarin
Prey species sit at a pivotal point in food webs, serving as a connection between predators and
energy sources (e.g., plants or detritus). Most prey face multiple predators and must integrate
information about predation risk if they are to avoid being consumed. Meanwhile, predators interact
with one another and can increase or decrease their combined pressure on prey. By interacting with
their prey, predators can indirectly affect ecosystem functions, even without reducing prey population
size. The goal of my dissertation was to understand how prey survive in a world with multiple
predators and to uncover linkages between predators and the soil food web. I first tested hypotheses
about how the wolf spider Pardosa milvina responds to cues from multiple predators (the larger wolf
spider Tigrosa helluo and the ground beetle Scarites quadriceps) and how inaccurate information
regarding predation threat affects survival. Pardosa were capable of distinguishing between predators
and responding adaptively, though prey responses were not optimized when predators were at elevated
hunger levels. As a second step, I allowed multiple predators (the wolf spider Rabidosa rabida along
with Tigrosa and Scarites) to freely interact with each other and their prey (Pardosa) to test the
influence of predator characteristics and the occurrence of intraguild predation on prey survival.
Overall, I found support for a predictive framework of emergent multiple predator effects, though
intraguild predation events caused significant deviations from model predictions. I also investigated the
consumptive and nonconsumptive effects predators can have on their environment, focusing on the
detrital food chain. The presence of either Pardosa or their cues impacted CO2 flux and soil nitrogen
content as mediated by the detritivore Sinella curviseta, suggesting indirect top-down control of
ecosystem function by predators. Finally, I tested the response of Sinella to cues indicating predation
risk to determine if changes in detritivore activity linked predators to ecosystem function. Sinella
responded innately to necromones but did not alter activity levels in the presence of Pardosa cues, even
after a conditioning period.
BEHAVIORAL AND ECOLOGICAL CONSEQUENCES OF MULTIPLE INTRAGUILD
PREDATORS AND CONNECTIONS BETWEEN PREDATORS, PREY, AND ECOSYSTEM
FUNCTION
A DISSERTATION
Submitted to the faculty of
Miami University in partial
fulfillment of the requirements
for the degree of
Doctor of Philosophy
Department of Biology
by
Michael Ian Sitvarin
Miami University
Oxford, Ohio
2014
Dissertation Director: Ann L. Rypstra
TABLE OF CONTENTS
General Introduction........................................................................................................................1
Chapter 1:
Patchy and Mismatched Cues: How Do Prey Respond To Multiple Predators
Representing Different Levels of Risk?................................................................9
Abstract..............................................................................................................................10
Introduction........................................................................................................................11
Methods..............................................................................................................................13
Results................................................................................................................................16
Discussion..........................................................................................................................18
References..........................................................................................................................21
Tables & Figures …...........................................................................................................25
Chapter 2:
The Importance of Intraguild Predation in Predicting Emergent Multiple
Predator Effects....................................................................................................33
Abstract..............................................................................................................................34
Introduction........................................................................................................................35
Methods..............................................................................................................................37
Results................................................................................................................................41
Discussion..........................................................................................................................43
References..........................................................................................................................48
Tables & Figures …...........................................................................................................54
Chapter 3:
Fear of Predation Alters Soil CO2 Flux and Nitrogen Content........................60
Abstract..............................................................................................................................61
Introduction........................................................................................................................62
Methods..............................................................................................................................63
Results................................................................................................................................64
Discussion..........................................................................................................................65
ii
References..........................................................................................................................67
Tables & Figures …...........................................................................................................70
Chapter 4:
Nonconsumptive Predator-Prey Interactions: Sensitivity of a Detritivore to
Cues of Predation Risk........................................................................................74
Abstract..............................................................................................................................75
Introduction........................................................................................................................76
Methods..............................................................................................................................78
Results................................................................................................................................81
Discussion..........................................................................................................................83
References..........................................................................................................................86
Tables & Figures …...........................................................................................................91
General Conclusion and Future Directions....................................................................................99
References........................................................................................................................105
Appendix......................................................................................................................................107
Chapter 1: Supplementary Material.................................................................................108
Chapter 2: Supplementary Material.................................................................................109
Chapter 3: Supplementary Material.................................................................................113
Chapter 4: Supplementary Material.................................................................................129
iii
LIST OF TABLES
Chapter 1
Table 1. Loading of activities (from Videomex-V software) on principal components.
Positive and negative values indicate positive and negative correlations with the
principal component, respectively. Magnitudes indicate the strength of correlation
between the activity variable and the principal component.
Table 2. Effects of treatment (predator cues present in arena), predator hunger level, and
their interaction on Pardosa activity. Model degrees of freedom: 7, 150.
Table 3. Effects of treatment (predator cues present in arena) and predator hunger level on
Pardosa activity. Statistics reported are Cohen's d, 95% confidence intervals, and
F-values, degrees of freedom, and p-values from one-way ANOVAs. Treatments
are blank (B), cues from Tigrosa (T), and cues from Scarites (S). Symbols
between treatment letters indicate relationships based on effect sizes.
Table 4. Proportional hazards test of the effects of treatment (cues present in arena),
predator hunger level, and their interaction on Pardosa survival.
Table 5. Effects of treatment (cues present in arena) and predator hunger level on
Pardosa mortality. Treatments are blank (B), cues from Tigrosa (T), and cues
from Scarites (S). Symbols between treatment letters indicate whether risk was
increased, decreased, or unaffected as determined by the hazard ratio (i.e.,
instantaneous probability of death).
Chapter 2
Table 1. Expected and observed multiple predator effects (MPEs) on Pardosa survival
due to predation by Tigrosa (T), Rabidosa (R), and Scarites (S). Tests for MPEs
were also analyzed by excluding trials with intraguild predation (IGP: here
defined as predation between Tigrosa, Rabidosa, and Scarites).
Table 2. Test for multiple predator effects (MPEs) on Pardosa survival due to predation
by Tigrosa (T), Rabidosa (R), and Scarites (S) using logistic regression. Analyses
were also conducted by excluding trials with intraguild predation (IGP: here
defined as predation between Tigrosa, Rabidosa, and Scarites).
Table 3. Impact of Tigrosa (T), Rabidosa (R), and Scarites (S) on the frequency with
iv
which another predator consumed Pardosa. Values reported are regression
coefficients (p-values). Regression coefficients represent the change in the
response (e.g., frequency of Tigrosa consuming Pardosa) given one unit change in
a predictor (e.g., presence of Rabidosa). Analyses were also conducted by
excluding trials with intraguild predation (IGP: here defined as predation between
Tigrosa, Rabidosa, and Scarites).
Chapter 3
Table 1. Effects of cues and predation on the survival of detritivores. Treatments: cues
(C), predation (P), detritivore (D). Symbols between treatment letters indicate
relationships based on effect sizes.
Table 2. Effects on corrected CO2 flux and soil N content. Treatments: blank (B), cues
(C), predation (P), detritivore (D). Symbols between treatment letters indicate
relationships based on effect sizes.
Chapter 4
Table 1. Loading of activities on principal components and proportion of variation
explained by each component for the response to spider cues in the first
experiment.
Table 2. Loading of activities on principal components and proportion of variation
explained by each component for the necromone experiment.
Table 3. Loading of activities on principal components and proportion of variation
explained by each component for the response to spider cues prior to conditioning.
Table 4. Loading of activities on principal components and proportion of variation
explained by each component for the response to spider cues after conditioning.
Appendix
Chapter 1
Table A1. Mean (SE) for each activity variable used in the principal component analysis
of Pardosa response to patchy cues from Tigrosa and Scarites at two hunger
levels.
Chapter 2
Table A2. Summary of habitat domain and hunting mode for Pardosa, Tigrosa,
v
Rabidosa, and Scarites. Habitat domain and hunting mode classified according to
supplementary results from chapter 2, personal observations, and published
literature.
Chapter 3
Table A3. Treatment effects on unmanipulated CO2 flux dynamics (repeated measures
ANOVA).
Table A4. Treatment effects on unmanipulated CO2 flux from the last day of the
experiment. Statistics reported are Cohen's d, 95% confidence intervals, and Fvalues, degrees of freedom, and p-values. Treatments are: blank (B), cues (C),
predation (P), detritivore (D). Symbols between treatment letters indicate
relationships based on effect sizes.
Table A5. Treatment effects on unmanipulated soil N content. Statistics reported are
Cohen's d, 95% confidence intervals, and F-values, degrees of freedom, and pvalues. Treatments are: blank (B), cues (C), predation (P), detritivore (D).
Symbols between treatment letters indicate relationships based on effect sizes.
Table A6. Treatment effects on soil C content. All tests were performed on unmanipulated
and corrected values (see methods). Statistics reported are Cohen's d, 95%
confidence intervals, and F-values, degrees of freedom, and p-values. Treatments
are: blank (B), cues (C), predation (P), detritivore (D). Symbols between treatment
letters indicate relationships based on effect sizes.
Table A7. Treatment effects on soil organic C content. All tests were performed on
unmanipulated and corrected values (see methods). Statistics reported are Cohen's
d, 95% confidence intervals, and F-values, degrees of freedom, and p-values.
Treatments are: blank (B), cues (C), predation (P), detritivore (D). Symbols
between treatment letters indicate relationships based on effect sizes.
Table A8. Treatment effects on soil C:N. Statistics reported are Cohen's d, 95%
confidence intervals, and F-values, degrees of freedom, and p-values. Treatments
are: blank (B), cues (C), predation (P), detritivore (D). Symbols between treatment
letters indicate relationships based on effect sizes.
Chapter 4
vi
Table A9. Loadings of activity metrics on principal components and proportion of total
variation explained by each component.
Table A10. Mean (SE) for each activity variable used in the principal component analysis
of Sinella response to cues from Pardosa as well as the response to necromones.
Each cue source (Pardosa cues or necromones) is paired with a blank.
Table A11. Mean (SE) for each activity variable used in the principal component analysis
of Sinella response to cues from Pardosa before and after conditioning. Each cue
source is paired with a blank.
vii
LIST OF FIGURES
Chapter 1
Figure 1. Diagram of arena layouts representing the four treatments used in the patchwork
activity experiment. Filter paper quadrants were blank (B) or previously occupied
by Tigrosa (T) or Scarites (S).
Figure 2. Box plots of principal components from the patchwork activity experiment for
cues from predators at low and high hunger levels. Treatments are blank (B), cues
from Tigrosa (T), Scarites (S), or both Tigrosa and Scarites (T&S). Box plots
show median, first and third quartiles, greatest values within 1.5 interquartile
range, and outliers. Different letters indicate significant differences following
Tukey HSD tests.
Figure 3. Pardosa survival in an arena with filter paper that was unmanipulated (blank) or
previously occupied by either Tigrosa or Scarites. The predator with which
Pardosa interacted directly is pictured. Note the differences in time scale between
predators. Censored observations are represented by a plus symbol.
Chapter 2
Figure 1. Habitat domain and hunting mode of study species. Arrows point from predator
to prey. The inset figure depicts the spatial relationships of Pardosa (rectangle)
and Tigrosa, Rabidosa, and Scarites (ovals). The lack of domain overlap between
Tigrosa and Rabidosa is predicted to lead to substitutable risk. The combination of
overlapping domains and different hunting modes for Tigrosa and Scarites is
predicted to create risk reduction by means of intraguild predation. Rabidosa and
Scarites have non-overlapping domains and should create substitutable risk.
Figure 2. Proportion of trials in which Pardosa was consumed, summed across predators
(A) and per predator (B-D). Including trials with predation between Tigrosa (T),
Rabidosa (R), and Scarites (S). Fitted lines illustrate trends in predation success
(see discussion).
Figure 3. Proportion of trials in which Pardosa was consumed, summed across predators
(A) and per predator (B-D). Excluding trials with predation between Tigrosa (T),
Rabidosa (R), and Scarites (S). Fitted lines illustrate trends in predation success
viii
(see discussion).
Chapter 3
Figure 1. Corrected CO2 flux dynamics (mean + SE).
Figure 2. Corrected total CO2 flux on the last day of the experiment (A) and soil nitrogen
(B). Box plots show median, first and third quartiles, greatest values within 1.5
interquartile range, and outliers.
Chapter 4
Figure 1. Collembolan activity (blank – cue) in response to spider cues in the first
experiment. Box plots show median, first and third quartiles, greatest values
within 1.5 interquartile range, and outliers.
Figure 2. Collembolan activity (blank – cue) in response to necromones. Box plots show
median, first and third quartiles, greatest values within 1.5 interquartile range, and
outliers.
Figure 3. Collembolan activity (blank – cue) in response to spider cues prior to
conditioning for control (A) and experimental (B) groups. Box plots show median,
first and third quartiles, greatest values within 1.5 interquartile range, and outliers.
Figure 4. Collembolan activity (blank – cue) in response to spider cues after conditioning
for control (A) and experimental (B) groups. Box plots show median, first and
third quartiles, greatest values within 1.5 interquartile range, and outliers.
Appendix
Chapter 2
Figure A1. Characterization of habitat domain for Pardosa, Tigrosa, Rabidosa, and
Scarites (mean + SE).Vertical displacement and vertical habitat use loaded
positively on PC1, whereas use of the soil surface loaded negatively (a).
Horizontal displacement and use of the subterranean habitat loaded positively on
PC2 (b).
Chapter 3
Figure A2. Relationship between detritivores consumed by predators and total CO2 flux
on the last day of the experiment.
Figure A3. Unmanipulated CO2 flux dynamics (mean +SE).
ix
Figure A4. Unmanipulated CO2 flux on the last day of the experiment. Box plots show
median, first and third quartiles, greatest values within 1.5 interquartile range, and
outliers.
Figure A5. Soil C content for both unmanipulated (A) and corrected values (B). Box plots
show median, first and third quartiles, greatest values within 1.5 interquartile
range, and outliers.
Figure A6. Soil organic C content for both unmanipulated (A) and corrected values (B).
Box plots show median, first and third quartiles, greatest values within 1.5
interquartile range, and outliers.
Figure A7. Soil C:N for both unmanipulated (A) and corrected values (B). Box plots
show median, first and third quartiles, greatest values within 1.5 interquartile
range, and outliers.
Chapter 4
Figure A8. Collembolan activity in response to cues from Pardosa (grey), Tigrosa
(brown), Rabidosa (yellow), and Scarites (black). Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
Figure A9. Collembolan activity in response to cues from Pardosa (grey), Tigrosa
(brown), Rabidosa (yellow), and Scarites (black). Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
Figure A10. Collembolan activity in response to cues from Pardosa (grey), Tigrosa
(brown), Rabidosa (yellow), and Scarites (black). Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
x
DEDICATION
To all the spiders, beetles, collembolans, and crickets who unwillingly gave their lives in the
name of science. They are gone, but not forgotten.
xi
ACKNOWLEDGMENTS
I could not have completed this work alone, so I have many people to thank. First and foremost, I
thank my advisor, Ann Rypstra, for all of the time and effort she put into guiding my progress.
My committee members (Nancy Solomon, Brian Keane, Tom Crist, and Dave Gorchov)
provided valuable insight and feedback through the years. My fellow “spiderfans”, Kerri Wrinn,
Jason Schmidt, Jonathan Edwards, Meg Marchetti, Lacey Campbell, Khalid Mukhtar, and James
Harwood supported me in my efforts and discussed lots of data with me. By my count there were
38 undergraduates who helped me by keeping the lab running during my stay here, and their
contributions cannot be overstated. The undergraduates with whom I collaborated on various
projects (Kelsey Breen, Christian Romanchek, Catherine Hoffman, Shan “froot fly” Qureshi, AJ
Norton, Alex Webb) get a special thanks for sharing my passion for research. Thanks to all of the
wonderful graduate students I've met here as well as non-science friends, who provided
perspective on life. My loving parents and siblings have always believed in me, even if they had
no idea what I was actually doing. Finally, none of this would have been possible without the
love and support of my wife, Ann Showalter.
xii
General Introduction
1
The role of predators in communities is complex and precludes simple predictions of their
impacts on the surrounding environment. This realization has come after decades of early studies
focused on understanding the role of predation in shaping prey behavior, communities, and ecosystems.
In the field of animal behavior, researchers have made considerable progress in revealing the
mechanisms by which prey detect and respond to information indicating a predation risk. Though
publication bias likely influences our understanding, prey are typically able to minimize predation risk
and increase survival in the presence of predators, indicating widespread evolution of adaptive
behaviors. However, most studies focus on a single prey species interacting with a single predator
species. This is a practical approach that facilitates manageable experimental designs, though a crucial
aspect of realism is missed: prey are consumed by multiple predator species in nature. Incorporation of
food web studies, which had long-recognized the existence of multiple predators, lead to a field of
ecology centered on emergent multiple predator effects. This body of work quickly outstripped simple
theory that paralleled the biodiversity-ecosystem function literature, mainly focused on terrestrial plants
and primary productivity. The field has since matured, with hundreds of examples, trait-based
predictive frameworks, and comprehensive reviews of the subject. However, the majority of studies
have been conducted on a fairly limited collection of organisms and habitat types, leaving room for
future researchers to make substantial progress.
Top-down effects of predators on ecosystems also has a long history with a focus that has
shifted over time as studies have become more nuanced, less phenomenological, and more mechanismoriented. Early experiments that utilized predator exclusion treatments revealed stark differences in
community composition when compared to unmanipulated controls. Intuitively, predator consumption
of prey was considered the mechanism driving these effects, which fit well with population models of
predator-prey dynamics. Aquatic, and later, terrestrial, studies revealed that predator-prey dynamics
could cascade through multiple trophic levels such that predators were understood to indirectly affect
primary producer biomass. This revelation spawned countless papers on the phenomenon of trophic
cascades, improving the ability of ecologists to understand how ecosystems function. However, a
separate body of literature on phenotypic plasticity, more specifically, predator-induced changes in prey
behavior and morphology, had yet to cross-pollinate with the concepts central to trophic cascades. As
these two fields began to integrate, researchers gained insight into the ways predators could affect
systems without reducing prey population sizes. Nonconsumptive effects were shown to be pervasive
and often as strong or stronger than consumptive effects, forcing ecologists to reconsider the
2
mechanisms underlying previously observed phenomena in predator exclusion treatments. The idea that
prey “fear” their predators and respond in ways that cascade to lower trophic levels has been
thoroughly explored using herbivorous pathways: predators affects prey that in turn affect either plants
or phytoplankton. However, researchers are only just beginning to apply the lessons learned from
“green” food webs to “brown” food webs; cascading predator effects that ultimately impact detritus are
likely widespread and strong, but we are short on data.
The following dissertation describes a set of experiments designed to understand how a suite of
predators affect their shared prey and how predators may be indirectly connected to the soil food web
by means of both direct consumption of prey and induced changes in prey behavior or physiology that
have cascading impacts on ecosystem function. My approach integrates multiple distinct, yet related,
concepts in ecology, including: 1) predation cue-mediated behavioral plasticity in prey (Relyea 2003),
2) predator habitat domain and hunting mode (Schmitz 2007), 3) emergent multiple predator effects
(Sih et al. 1998), 4) intraguild predation (Polis et al. 1989, Vance-Chalcraft et al. 2007), 5) consumptive
and nonconsumptive effects of predators on prey (Preisser et al. 2005), 6) trophic cascades (Pace et al.
1999), and 7) the relationship between biodiversity and ecosystem function (Duffy 2002, Hooper et al.
2005). While these topics are commonly addressed individually in the ecological literature, they are
infrequently used in combination to understand how organisms interact with each other and their
environment. Furthermore, my study system possesses the following characteristics, which are underrepresented in this field of research: 1) the suite of predators engage in reciprocal intraguild predation,
2) the focal prey species shared by multiple predators is itself a predator, and 3) the trophic cascade
focuses on a detrital system instead of an herbiviory-based system, providing an example of linkage
between above- and below-ground environments.
In chapter 1 I investigate how prey respond to cues of predation risk when there are multiple
predators. By using two predators that differ in hunting mode, activity level, and predation risk, I
provide insight into how prey prioritize potentially conflicting or inaccurate cues in terms of activity
and survival. Furthermore, I manipulated the hunger level of the predators to determine whether
prioritization of cues is altered by a perceived overall increase in risk of predation. Chapter 2 expands
upon the first by adding another predator and, importantly, allowing the predators to interact with one
another. I quantified important characteristics of the predators and prey to build a predictive model of
how their interactions would ultimately impact prey survival. By dividing the data I collected according
to trials in which the top predators did and did not consume one another, I gained valuable insight into
3
the role of intraguild predation in determining the nature of emergent multiple predator effects. The
remaining chapters focus on the connection between above-ground predators and the soil food web. In
chapter 3, I used laboratory microcosms to evaluate the presence of trophic cascades in a detrital
system as propagated by purely nonconsumptive effects and by whole predator effects (i.e., including
both consumptive and nonconsumptive effects). Finally, in chapter 4 I return to a focus on animal
behavior by testing for changes in activity of a detritivore in response to predator cues. The goal of this
chapter was to provide a mechanism to explain the results found in chapter 3. This chapter also
includes a conditioning component meant to more fully understand the nature of interactions between
predators and detritivores. I conclude by describing both what has been gained from my studies and the
avenues of research extending from my work that remain to be explored.
Study system
The wolf spider Pardosa milvina (Araneae: Lycosidae) is a generalist predator common
throughout eastern North America that can achieve densities exceeding 40 individuals/m2 on the soil
surface of agricultural fields in southwest Ohio (Marshall et al. 2000). Pardosa is a fairly active
predator that deploys a sit-and-move predation strategy (Lowrie 1973, Ford 1978, Nyffeler et al. 1994,
Walker et al. 1999a, Samu et al. 2003). It co-occurs with a suite of larger arthropod predators including
the wolf spiders Tigrosa (formerly Hogna (Brady 2012)) helluo and Rabidosa rabida in addition to the
beetle Scarites quadriceps (Coleoptera: Carabidae) (Young & Edwards 1990, McNabb et al. 2001).
Both Tigrosa and Rabidosa are sit-and-move predators, though Tigrosa is largely restricted to the soil
surface where females are facultative burrowers (Walker et al. 1999b), whereas Rabidosa is typically
found in elevated vegetation (Brady & McKinley 1994). Scarites frequently burrows, but emerges at
night when it actively searches for prey on the soil surface (Lundgren et al. 2009, personal
observation). Pardosa is more active during the day, (Marshall et al. 2002, Schonewolf et al. 2006), but
is likely to encounter Tigrosa, Rabidosa, and Scarites at night, when these predators are more active
(Lizotte & Rovner 1988, Brady & McKinley 1994, Marshall et al. 2002, Lundgren et al. 2009).
Although Pardosa is mostly active on the soil surface, interactions with Tigrosa and Scarites can
induce climbing behavior that brings Pardosa into contact with Rabidosa (Lowrie 1973, Folz et al.
2006). All three of these predators engage in intraguild predation (Polis et al. 1989) with Pardosa with
some frequency. Collembola are frequent prey of spiders (Kajak 1995, Buddle 2002, Kuusk & Eckbom
2010) and are sensitive to chemical cues in their environment. The collembolan Sinella curviseta
4
(Entomobryomorpha: Entomobryidae) is a widespread species active on the soil surface (Waldorf 1971,
Hopkin 1997), where it likely encounters epigeal spiders.
5
REFERENCES
Brady AR, McKinley KS. 1994. Nearctic species of the wolf spider genus Rabidosa (Araneae,
Lycosidae). Journal of Arachnology 22, 138-160
Brady AR. 2012. Nearctic species of the new genus Tigrosa (Araneae: Lycosidae). Journal of
Arachnology 40, 182-208.
Buddle CM. 2002. Interactions among young stages of the wolf spiders Pardosa moesta and P.
mackenziana (Araneae: Lycosidae). Oikos 96, 130-136 (doi:10.1034/j.16000706.2002.960114.x)
Duffy JE. 2002. Biodiversity and ecosystem function: the consumer connection. Oikos 99, 201-219
(doi:10.1034/j.1600-0706.2002.990201.x)
Folz HC, Wilder SM, Persons MH, Rypstra AL. 2006. Effect of predation risk on vertical habitat use
and foraging of Pardosa milvina. Ethology 112, 1152-1158 (doi:10.1111/j.14390310.2006.01276.x)
Ford MJ. 1978. Locomotory activity and the predation strategy of the of the wolf-spider Pardosa
amentata (Clerck) (Lycosidae). Animal Behaviour 26, 31-35 (doi:10.1016/00033472(78)90005-2)
Hooper DU, Chapin III FS, Ewel JJ, Hector A, Inchausti P, Lavorel S, Lawton JH, Lodge DM, Loreau
M, Naeem S, Schmid B, Setälä H, Symstad AJ, Vandermeer J, Wardle DA. 2005. Effects of
biodiversity on ecosystem functioning: a consensus of current knowledge. Ecological
Monographs 75, 3-35 (doi:10.1890/04-0922)
Hopkin SP. 1997. Biology of the springtails (Insecta: Collembola). Oxford University Press, Oxford.
Kajak A. 1995. The role of soil predators in decomposition processes. European Journal of
Entomology 92, 573-580
Kuusk A-K, Ekbom B. 2010. Lycosid spiders and alternative food: feeding behavior and implications
for biological control. Biological Control 55, 20-26 (doi:10.1016/j.biocontrol.2010.06.009)
Lizotte RS, Rovner JS. 1988. Nocturnal capture of fireflies by lycosid spiders – visual versus vibratory
stimuli. Animal Behaviour 36, 1809-1815 (doi:10.1016/S0003-3472(88)80120-9)
Lowrie DC. 1973. The microhabitats of western wolf spiders of the genus Pardosa. Entomological
News 84, 103-116
Lundgren JG, Nichols S, Prischmann DA, Ellsbury MM. 2009. Seasonal and diel activity patterns of
generalist predators associated with Diabrotica vigifera immatures (Coleoptera:
6
Chrysomelidae). Biocontrol Science and Technology 19, 327-333
(doi:10.1080/09583150802696533)
Marshall SD, Walker SE, Rypstra AL. 2000. A test for a differential colonization and competitive
ability in two generalist predators. Ecology 81, 3341-3349
Marshall SD, Pavuk DM, Rypstra AL. 2002. A comparative study of the phenology and daily activity
patterns in the wolf spiders Pardosa milvina and Hogna helluo in soybean agroecosystems in
southwestern Ohio (Araneae, Lycosidae). Journal of Arachnology 30, 503-510
(doi:10.1636/0161-8202(2002)030[0503:ACSOPA]2.0.CO;2)
McNabb DM, Halaj J, Wise DH. 2001. Inferring trophic positions of generalist predators and their
linkage to the detrital food web in agroecosystems: a stable isotope analysis. Pedobiologia 45,
289-297 (doi:10.1078/0031-4056-00087)
Nyffeler M, Sterling WL, Dean DA. 1994. How spiders make a living. Environmental Entomology 23,
1357-1367
Pace ML, Cole JJ, Carpenter SR, Kitchell JF. 1999. Trophic cascades revealed in diverse ecosystems.
Trends in Ecology and Evolution 14, 483-488 (doi:10.1016/S0169-5347(99)01723-1)
Polis GA, Myers CA, Holt RD. 1989. The ecology and evolution of intraguild predation: potential
competitors that eat each other. Annual Review of Ecology and Systematics 20, 297-330
(doi:10.1146/annurev.es.20.110189.001501)
Preisser EL, Bolnick DI, Benard ME. 2005. Scared to death? The effects of intimidation and
consumption in predator-prey interactions. Ecology 86, 501-509 (doi:10.1890/04-0719)
Relyea RA. 2003. How prey respond to combined predators: a review and an empirical test. Ecology
84, 1827-1839 (doi:10.1890/0012-9658(2003)084[1827:HPRTCP]2.0.CO;2)
Samu F, Sziranyi A, Kiss B. 2003. Foraging in agricultural fields: a 'sit-and-move' strategy scales up to
risk-averse habitat use in a wolf spider. Animal Behaviour 66, 939-947
(doi:10.1006/anbe.2003.2265)
Schmitz OJ. 2007. Predator diversity and trophic interactions. Ecology 88, 2415-2426 (doi:10.1890/060937.1)
Schonewolf KW, Bell R, Rypstra AL, Persons MH. 2006. Field evidence of an airborne enemyavoidance kairomone in wolf spiders. Journal of Chemical Ecology 32, 1565-1576
(doi:10.1023/A:1020878225553)
Sih A, Englund G, Wooster D. 1998. Emergent impacts of multiple predators on prey. Trends in
7
Ecology and Evolution 13, 350-355 (doi:10.1016/S0169-5347(98)01437-2)
Vance-Chalcraft HD, Rosenheim JA, Vonesh JR, Osenberg CW, Sih A. 2007. The influence of
intraguild predation on prey suppression and prey release: a meta-analysis. Ecology 88, 26892696 (doi:10.1890/06-1869.1)
Waldorf ES. 1971. The reproductive biology of Sinella curviseta (Collembola: Entomobryidae) in
laboratory culture. Revue d'Ecologie et de Biologia du Sol 8, 451-463
Walker SE, Marshall SD, Rypstra AL, Taylor DH. 1999a. The effects of hunger on locomotory
behavior in two species of wolf spider (Araneae, Lycosidae). Animal Behaviour 58, 515-520
(doi:10.1006/anbe.1999.1180)
Walker SE, Marshall SD, Rypstra AL. 1999b. The effect of feeding history on retreat construction in
the wolf spider Hogna helluo (Araneae, Lycosidae). Journal of Arachnology 27, 689-691
Young OP, Edwards GB. 1990. Spiders in United States field crops and their potential effect on crop
pests. Journal of Arachnology 18, 1-27
8
Chapter 1: Patchy and Mismatched Cues: How Do Prey Respond to Multiple Predators
Representing Different Levels of Risk?
9
ABSTRACT
Prey species exist within complex food webs and typically must contend with multiple
predators that vary in degree of predation threat they pose. Predators frequently deposit cues that can be
used by prey to detect, avoid, and discriminate between their various predators using adaptive
behaviors. I used a suite of arthropod predators to test hypotheses regarding cue prioritization and
integration by prey when confronted with multiple predators. I recorded changes in activity of the wolf
spider Pardosa milvina when presented with a patchwork of cues from its predators, the larger wolf
spider Tigrosa helluo and the ground beetle Scarites quadriceps, each tested at two different hunger
levels. I also quantified the adaptive value of how prey respond to predator cues by measuring survival
in the presence of each predator 1) without predator cues, 2) with cues matching the predator present,
and 3) with cues from the other predator. I found Pardosa largely prioritized its activity responses
toward the more dangerous predator, Tigrosa, though responses to both predators were similar when the
predators were at high hunger levels. In the survival trials, cues matching the predator present tended to
provide Pardosa a survival advantage, whereas mismatched cues typically increased the risk of being
consumed. I propose the existence of a strong chemotactile sensory bias in Pardosa that is imperfect
but likely beneficial overall by tuning responses to avoid detection and capture by predators posing the
greatest risk. Future efforts to understand prey behavior in response to the risk of predation should
incorporate the lack of complete information by prey and the existence of multiple predators that differ
in the level of threat posed.
10
INTRODUCTION
An essential component in the life of most animals is the avoidance of predation; however, this
seemingly simple behavior is complicated by the fact that prey must often contend with more than one
species of predator (Sih et al. 1998, Relyea 2003) and inaccurate information about predators (Lima &
Steury 2005, Ferrari et al. 2009). Interactions with multiple predators may be responsible for the
evolution and maintenance of complex antipredator behaviors (Blumstein 2006, Vencl & Srygley
2013). If prey are to use these antipredator behaviors adaptively, they must be able to recognize and
discriminate between different predators and initiate appropriate responses. Numerous studies have
confirmed the ability of prey to discriminate between different predator species and respond according
to the perceived risk of predation (Relyea 2003, Preisser et al. 2007).
Despite the demonstrated ability for prey to differentiate between predation threats, their
responses to a diversity of predators may not offer adequate protection against being consumed. A
conceptual framework outlined by Herzog & Laforsch (2013) classifies predators in terms of their
impacts on prey phenotypes. Predators that cause similar responses in prey (e.g., reduced activity) are
considered functionally equivalent because prey can respond effectively to both predators using a
generalized response (i.e., reducing activity levels). In contrast, functionally inverse predators elicit
mutually exclusive responses from prey (e.g., both increased and decreased activity). When presented
with predators selecting for opposing behaviors, prey can use either a hierarchical or compromise
strategy (McIntosh & Peckarsky 1999). Hierarchical responses are targeted toward a specific predator,
typically the one posing the greatest risk. Prey using compromise responses have behaviors that are
intermediate between those elicited by either predator alone.
Before prey can respond to the threat of predation from multiple predators, they must assess the
risk posed by each predator. Many animals rely on cues (e.g., kairomones) to evaluate the presence of
and threat posed by predators (Dicke & Grostal 2001). In a test of the adaptive plasticity hypothesis,
Hoverman & Relyea (2009) exposed snails to cues from predatory water bugs and crayfish as a means
of inducing predator-specific shell morphologies. Allowing snails to interact with a predator that did
not match the cues used to induce its shell morphology revealed survival trade-offs; prey responses
were maladaptive, representing what likely happens in nature when prey must evaluate cues from
predators differing in risk.
Factors such as the hunting mode (Preisser et al. 2007) and hunger level (Bell et al. 2006) of
predators are known to influence the strength of prey responses, as sit-and-move and hungry predators
11
represent a higher risk of predation than active and satiated predators, respectively (chapter 2).
However, terrestrial systems are under-represented in studies of prey responses to predator cues
(Preisser & Bolnick 2008), as amphibians, molluscs, and aquatic invertebrates are the most common
research subjects (e.g., Beckerman et al. 2010, Hettyey et al. 2011, Naddafi & Rudstam 2013).
Furthermore, most prey used in multiple predator experiments are not predatory, so lessons learned
from studying herbivores may not apply directly to predators (Paterson et al. 2013).
I evaluated changes in prey activity and survival in the presence of patchy and mismatched cues
from two predators that differ in hunting mode using a suite of coexisting terrestrial intraguild
predators (i.e., competitors that consume each other). Additionally, I manipulated predator hunger level
to test prey responses to perceived increases in the threat of predation. I predicted prey to exhibit a
hierarchical activity response biased towards cues from the more dangerous predator as the predators
used are functionally inverse (Herzog & Laforsch 2013, chapter 2), and for the strength of the response
to increase with elevated predator hunger level. Additionally, I expected a mismatch between the
predator and cues present to lead to decreased survival due to the implementation of inappropriate
antipredator behavior.
12
METHODS
Study System
Pardosa milvina is a small wolf spider that coexists with the larger wolf spider Tigrosa helluo
and the ground beetle Scarites quadriceps (all species hereafter referred to by genus). Tigrosa and
Scarites are predators of Pardosa and differ in the threat posed to Pardosa: Tigrosa frequently captures
Pardosa using a sit-and-move hunting mode, whereas Scarites infrequently captures Pardosa and is an
active hunter (Chapter 2). Pardosa is sensitive to chemotactile cues (silk, feces, and other excreta;
hereafter referred to as cues) from these predators (Wrinn et al. 2012) and is capable of displaying
graded antipredator behavior in response to the level of risk indicated by predator cues (Persons &
Rypstra 2001, Lehmann et al. 2004, Bell et al. 2006).
Collection and Maintenance
All study species were collected in and around agricultural fields at Miami University's Ecology
Research Center (39°31′33′′ N, 84°43′20′′W) and maintained in an environmental chamber on a
13L:11D light cycle at 25°C. Only adult female Pardosa and adult Scarites were used, though Tigrosa
were either penultimate or adult females. Pardosa were housed individually in plastic containers
(5.5cm high x 5.5cm diameter) with a moistened layer of potting soil and peat moss (1:1 mixture).
Tigrosa and Scarites were housed individually in larger containers (8cm tall x 12cm diameter) with the
same substrate. All species were provided two domestic house crickets (Acheta domesticus)
approximately one-half of the organism’s body size weekly and 48h before being used in experiments.
For trials at a high hunger level, Tigrosa and Scarites were provided three crickets before being
withheld food three weeks prior to be used as cue sources or predators. No organisms or cues were used
more than once in activity (low hunger: n=20, high hunger: n=17-21) or survival (low hunger: n=20,
high hunger: n=18-20) trials, and all arenas were cleaned with ethanol and allowed to dry prior to use.
Activity – Patchy Cues
To evaluate how prey integrate information about the risk of predation from multiple predators,
I quantified Pardosa activity in response to a patchwork of predator cues. Tigrosa and Scarites were
maintained on filter paper for 24h prior to trials to collect cues. Four equal-sized, quarter-circle pieces
of filter paper were placed into a plastic arena (7.5cm high, 20cm diameter) lined with untreated filter
paper. Quadrants were separated from each other by 1cm with a 2cm diameter space in the center of the
13
arena. Each quadrant was either blank or previously occupied by Tigrosa or Scarites, and arranged in
an alternating pattern in an additive design that standardized the amount of cue from each predator
(Figure 1). Pardosa were introduced into the center of arena under a clear glass vial and released after a
1min acclimation period.
I monitored Pardosa activity remotely using a camera mounted 1m above the arena and
automatic motion-tracking software (Videomex-V, Columbus Instruments, Columbus, OH, USA). The
following activities were recorded for 30min: 1) distance traveled (cm), 2) time spent ambulatory (s,
movement of one body length or more per second), 3) time spent in stereotypic motion (s, movement of
less than one body length per second), and 4) time spent immobile (s). Speed (cm/s) was calculated as
the distance traveled divided by the total time spent in ambulatory and stereotypic motion.
Survival – Mismatched Cues
To understand the consequences of the behavioral shifts observed in the previous experiment, I
examined Pardosa survival in situations where predator cues did not match the predator present. I
allowed a single Pardosa to interact with a single predator (Tigrosa or Scarites) in a plastic arena
(7.5cm tall, 20cm diameter) lined with filter paper that was either untreated or previously occupied by
Tigrosa or Scarites for 24h prior to the trial. I placed opaque vials over the study species and allowed
them to acclimate for 5min prior to being released at opposite sides of the arena. Pardosa survival was
monitored over 60h; arenas were checked every 12h except for the first 3h when checks were made at
5min intervals.
Statistical Analyses
The activity metrics from the motion-tracking software were combined using a principal
components analysis. Components with an eigenvalue greater than one were retained (Abdi & Williams
2010) and I used two-way ANOVA to examine the effect of treatment (cues present in arena), predator
hunger level (low or high), and their interaction on the principal components. Two-way ANOVAs were
conducted at each hunger level to determine whether the predator cues had interactive effects on
Pardosa behavior, as indicated by a significant interaction term (i.e., Tigrosa*Scarites, Schmitt et al.
2009). I also used one-way ANOVAs and Tukey's HSD tests to compare treatment effects on each
principal component within hunger levels. Additionally, I used effect sizes (Cohen's d) and confidence
intervals to evaluate treatment effects as this approach compliments traditional null hypothesis testing
14
and facilitates interpretation of significance in the biological and statistical sense (Garamszegi et al.
2009, Wesner et al. 2012). I follow suggested interpretations of effect size (small = 0.2, medium = 0.5,
large = 0.8; Cohen 1988). I used a proportional hazards test to determine the role of treatment (cues
present in arena), predator hunger level (low or high), and their interaction on Pardosa survival for
trials with Tigrosa and Scarites as predators. Hazard ratios (i.e., instantaneous probability of death)
were interpreted as significant based on the presence of large (i.e., greater than 40%) changes in risk
and p-values are reported for reference. All analyses were carried out using JMP (version 9.0; SAS
Institute, Inc., Cary, NC, USA) or R (R Core Team 2013).
15
RESULTS
Activity – Patchy Cues
The five activity metrics were summarized by two principal components, PC1 and PC2,
accounting for 70% and 24% of the total variation, respectively. PC1 distinguished activity from
inactivity, whereas PC2 primarily reflected changes in Pardosa speed (Table 1). Predator cue type
significantly affected both components, but only PC2 was significantly impacted by predator hunger
level and the interaction between cue type and hunger level (Table 2). When cue sources were at a low
hunger level, Pardosa decreased activity in arenas with cues from Tigrosa, but showed no significant
changes when encountering Scarites cues alone (Figure 2, Table 3). In contrast, Pardosa only reduced
their speed when cues from both predators were present. When cue sources were at a high hunger level,
Pardosa exhibited similar reductions in activity with a slightly stronger response to Scarites cues,
though the decrease in speed was of a greater magnitude and occurred in all predator cue treatments
(Figure 2, Table 3).
Separate two-way ANOVAs revealed that the two types of predator cues had an interactive
effect on PC2 at low (Tigrosa cue * Scarites cue interaction: p = 0.031) and high (p <0.001) hunger
levels, but there were no interactive effects on PC1 (low hunger: p = 0.997, high hunger: p = 0.135).
Survival – Mismatched Cues
Tigrosa and Scarites differed markedly in the predation risk they posed (Figure 3). Pardosa
were consumed by Tigrosa in 100% of the trials, and the majority of captures occurred within the first
hour (mean time until capture + SE: 122.7min + 21.8). In contrast, Scarites were successful in only
21.6% of the trials, and captured Pardosa survived over ten times longer (1432.7min + 189.6) than
those in trials with Tigrosa. Because of this difference between Tigrosa and Scarites, I ran a separate
proportional hazards model for each predator.
For trials with Tigrosa as the predator, Pardosa survival was influenced by the predator cues in
the arena and predator hunger level, but not the interaction between these factors (Table 4).
Specifically, cues from Tigrosa tended to improve survival whereas cues from Scarites did not
substantially alter the risk of predation (Table 5). Pardosa survival was significantly greater when the
predator cue matched the predator present, though this relationship was only apparent at the high
hunger level. Because previous studies have indicated strong responses to cues from Tigrosa (Persons
et al. 2001, Rypstra et al. 2007), I pooled blank and Scarites cues treatments and compared them to the
16
treatment with matched cues. The pooled test showed that both increased hunger level (hazard ratio:
0.52, 95% CI: 0.34-0.78, p = 0.002) and the presence of Tigrosa cues (hazard ratio: 0.59, 95% CI: 0.380.88, p = 0.009) significantly enhanced the odds of Pardosa survival.
Unlike trials with Tigrosa, the likelihood of Pardosa surviving a trial with Scarites depended on
the interaction between the type of cues present in the arena and predator hunger level (Table 4).
Specifically, Tigrosa cues had no effect on Pardosa survival in low hunger trials but enhanced risk in
high hunger trials (Table 5). Similarly, cues from Scarites in high hunger trials significantly enhanced
the risk of predation, though at low hunger levels these cues tended to increase Pardosa survival.
Although not statistically significant, mismatched cues tended to decrease Pardosa survival in low
hunger trials and moderately increase survival in high hunger trials (Table 5).
17
DISCUSSION
I have demonstrated how a prey species responds to cues from multiple predators representing
different threat levels related to their hunting mode and hunger level. Specifically, prey were capable of
integrating information present in the form of a patchwork of predator cues to alter their activity levels.
These changes in activity likely underlie trends in prey survival when confronted with a predator that
may not match the predator cues present. Interestingly, increasing predator hunger level strengthened
the response to one predator while reversing the response to the other, ultimately increasing the risk of
predation in the latter case.
Pardosa activity indicated that these predators may be functionally inverse (Herzog & Laforsch
2013), though the response to cues from Scarites was weak. When cues from both predators were
present, activity responses resembled those seen when only cues from Tigrosa were present, creating a
hierarchical response (McIntosh & Peckarsky 1999) towards the more dangerous predator. The use of a
hierarchical response is likely adaptive considering Tigrosa is the more dangerous predator (Chapter 2,
this study), uses a sit-and-move hunting mode that tends to produce more reliable cues (Preisser et al.
2007), and is unlikely to be deterred by post-contact defenses (McIntosh & Peckarsky 1999).
Conversely, only weak increases in activity to Scarites cues were evident, consistent with a previous
study (Wrinn et al. 2012) and the fact that Scarites infrequently consumes Pardosa (Chapter 2, this
study).
I observed an intriguing response to cues from predators at increased hunger levels; activity
reductions on Tigrosa cues were strengthened, but Scarites cues elicited decreased activity. Bell et al.
(2006) also showed greater activity reductions in response to Tigrosa at a higher hunger level, and it
seems that Pardosa may have a generalized response to cues from predatory arthropods that is only
evident when the predators have been deprived of food. The similar response to both predators when at
a high hunger level indicates that they are functionally equivalent and highlights the fact that the
designations developed by Herzog & Laforsch (2013) are not fixed, but can vary dynamically with
predator hunger level. Furthermore, when prey cannot discriminate between predator cues, it may be
advantageous to respond to the most likely or most risky predator (Brilot et al. 2012). The changes in
Pardosa activity in response to cues from both predators at two hunger levels underlie patterns of
survival in situations with predators on matched and mismatched cues.
When Pardosa faced a predator that matched the cues present at a low hunger level, more
individuals survived and for a longer period of time. These benefits likely arise from changes in
18
activity in response to predator cues. Specifically, reduced activity in the presence of sit-and-move
predators can be an effective means of avoiding capture (Persons et al. 2001), especially considering
that wolf spider visual systems are based on motion detection (Rovner 1996). In contrast, increases in
activity may help prey avoid capture by an active predator such as Scarites (Pruitt et al. 2012). When
predator cues did not match the predator present, Pardosa survival was comparable to the treatment
devoid of predator cues and individuals tended to be consumed more quickly. This lack of appropriate
response to the predator present indicates a strong chemotactile sensory bias, as visual, vibratory, and
olfactory cues would have been available to Pardosa, and Pardosa is capable of responding to these
other cue modalities (Schonewolf et al. 2006, Rypstra et al. 2009). In fact, chemotactile cues are likely
to be reliable indicators of predation risk whereas visual cues can be ambiguous, preventing prey from
responding appropriately (Brilot et al. 2012). Although integrating multimodal cues can help reduce
uncertainty, doing so may be prevented by a combination of physiological costs or phylogenetic
constraints (Munoz & Blumstein 2012).
More Pardosa survived for a longer time when Tigrosa were at a high hunger level. This
outcome is probably a combination of strengthened behavioral response of Pardosa to Tigrosa cues
(Bell et al. 2006, this study) and increased Tigrosa activity; Tigrosa become more active after a period
of food deprivation (Walker et al. 1999). This increase in Tigrosa movement likely interferes with
successful hunting, as the predator deviates from a sit-and-move hunting mode to a more active hunting
mode. While moving, Tigrosa may be more likely to fail to detect prey (Rovner 1996) while
simultaneously increasing the odds of being detected and consequently avoided by their prey due to
increased release of visual and vibratory cues. Although fewer Pardosa survived when confronting
Tigrosa on Scarites cues, the risk of predation was no different than when no cues were present.
Therefore, the slight reductions in activity seen in response to cues from Scarites at a high hunger level
were insufficient to significantly improve survival with a mismatched predator.
Interestingly, Pardosa confronted with Scarites at a high hunger level showed patterns of
survival in contrast with those from the lower hunger level trials. Specifically, spiders exposed to either
matched or mismatched cues were less likely to survive and were captured faster than those interacting
with Scarites without cues. The adaptive changes in prey activity in response to cues from Tigrosa at a
high hunger level (i.e., reduced activity) likely put Pardosa at a greater risk of being consumed by
Scarites, which is an active hunter. Reductions in Pardosa activity in response to cues from Scarites at
a high hunger level seem maladaptive, as survival was significantly worse than in treatments without
19
any cues. This maladaptive behavior only emerged at the elevated hunger level, suggesting Pardosa
may exhibit a generalized response to large hungry arthropod predators and further argues for a strong
chemotactile sensory bias in Pardosa. Persistence of a maladaptive antipredator response can be
explained if the behavior is effective (e.g., against Tigrosa) more frequently than it is ineffective, or if
the responses to both predators are genomically linked (Blumstein 2006). Additionally, this
maladaptive response may exist because Pardosa has a shorter co-evolutionary history with Scarites
than with Tigrosa. Scarites may only interact with Pardosa due to novel community assemblages in
response to human activity (i.e., agricultural fields). In contrast, Tigrosa and Pardosa naturally cooccur in riparian habitats, and thus Pardosa have the potential for a more finely-tuned response to
Tigrosa cues (e.g., Persons et al. 2001, Persons & Rypstra 2001, Barnes et al. 2002, Bell et al. 2006).
Models of animal behavior often assume prey have perfect information (e.g., Lima & Bednekoff
1999), though this is likely not often true in nature. Future modeling efforts that allow for inappropriate
prey responses due to incomplete or inaccurate assessment of predation threat would increase our
understanding of the pressures prey face when dealing with multiple predators. Furthermore, because
prey often have more than two predators, studies will benefit by examining a greater diversity of
predators, and help the body of terrestrial multiple predator studies catch up to those in aquatic systems
(Relyea 2003). Understanding behavioral responses to multiple predators is essential if we are to
predict emergent multiple predator effects (Sih et al. 1998), which are known to have cascading
impacts on prey survival, community structure, and ecosystem function (Finke & Denno 2004, Schmitz
2010, Steffan & Snyder 2010).
ACKNOWLEDGEMENTS
Numerous undergraduate and graduate students in my research group helped with animal
collection and maintenance. Financial support came from a Miami University Undergraduate Summer
Scholars Program award to Kelsey Breen.
20
REFERENCES
Abdi H, Williams LJ. 2010. Principal component analysis. Computational Statistics 2, 433-459
(doi:10.1002/wics.101)
Barnes MC, Persons MH, Rypstra AL. 2002. The effect of predator chemical cue age on antipredator
behavior in the wolf spider Pardosa milvina (Araneae: Lycosidae). Journal of Insect Behavior
15, 269-281 (doi:10.1023/A:1015493118836)
Beckerman AP, Rodgers GM, Dennis SR. 2010. The reaction norm of size and age at maturity under
multiple predator risk. Journal of Animal Ecology 79, 1069-1076 (doi:10.1111/j.13652656.2010.01703.x)
Bell RD, Rypstra AL, Persons MH. 2006. The effect of predator hunger on chemically mediated
antipredator responses and survival in the wolf spider Pardosa milvina (Araneae: Lycosidae).
Ethology 112, 903-910 (doi:10.1111/j.1439-0310.2006.01244.x)
Blumstein DT. 2006. The multipredator hypothesis and the evolutionary persistence of antipredator
behavior. Ethology 112, 209-217 (doi:10.1111/j.1439-0310.2006.01209.x)
Brilot BO, Bateson M, Nettle D, Whittingham MJ, Read JCA. 2012. When is general wariness favored
in avoiding multiple predator types? The American Naturalist 179, E180-E195
(doi:10.1086/665648)
Cohen J. 1988. Statistical power analysis for the behavioral sciences, second edition. Hillsdale, New
Jersey: Lawrence Erlbaum Associates.
Dicke M, Grostal P. 2001. Chemical detection of natural enemies by arthropods: an ecological
perspective. Annual Review of Ecology and Systematics 32, 1-23
(doi:10.1146/annurev.ecolsys.32.081501.113951)
Ferrari MCO, Sih A, Chivers DP. 2009. The paradox of risk allocation: a review and prospectus.
Animal Behaviour 78, 579-585 (doi:10.1016/j.anbehav.2009.05.034)
Finke DL, Denno RF. 2004. Predator diversity dampens trophic cascades. Nature 429, 407-410
(doi:10.1038/nature02554)
Garamszegi LZ, Calhim S, Dochtermann N, Hegyi G, Hurd PL, Jorgensen C, Kutsukake N, Lajeunesse
MJ, Pollard KA, Schielzeth H, Symonds MRE, Nakagawa S. 2009. Changing philosophies and
tools for statistical inferences in behavioral ecology. Behavioral Ecology 20, 1363-1375
(doi:10.1093/beheco/arp137)
Herzog Q, Laforsch C. 2013. Modality matters for the expression of inducible defenses: introducing a
21
concept of predator modality. BMC Biology 11, 113 (doi:10.1186/1741-7007-11-113)
Hettyey A, Vincze K, Zsarnóczai S, Hoi H, Laurila A. 2011. Costs and benefits of defences induced by
predators differing in dangerousness. Journal of Evolutionary Biology 24, 1007- 1019
(doi:10.1111/j.1420-9101.2011.02233.x)
Hoverman JT, Relyea RA. 2009. Survival trade-offs associated with inducible defences in snails: the
roles of multiple predators and developmental plasticity. Functional Ecology 23, 179-1188
(doi:10.1111/j.1365-2435.2009.01586.x)
Lehmann LM, Walker SE, Persons MH. 2004. The influence of predator sex on chemically mediated
antipredator response in the wolf spider Pardosa milvina (Araneae: Lycosidae). Ethology 110,
323-339 (doi:10.1111/j.1439-0310.2004.00972.x)
Lima SL, Bednekoff PA. 1999. Temporal variation in danger drives antipredator behavior: the
predation risk allocation hypothesis. The American Naturalist 153, 649-659
(doi:10.1086/303202)
Lima SL, Steury TD. 2005. Perception of predation risk: the foundation of nonlethal predator-prey
interactions. In: Ecology of predator-prey interactions. (Eds: Barbosa P and Castellanos I) 166188. Oxford: Oxford University Press.
McIntosh AR, Peckarsky BL. 1999. Criteria determining behavioural responses to multiple predators
by a stream mayfly. Oikos 85, 554-564 (doi:10.2307/3546705)
Munoz NE, Blumstein DT. 2012. Multisensory perception in uncertain environments. Behavioral
Ecology 23, 457-462 (doi:10.1093/beheco/arr220)
Naddafi R, Rudstam LG. 2013. Predator-induced behavioural defences in two competitive invasive
species: the zebra mussel and the quagga mussel. Animal Behaviour 86, 1275-1284
(doi:10.1016/j.anbehav.2013.09.032)
Paterson RA, Pritchard DW, Dick JTA, Alexander ME, Hatcher MJ, Dunn AM. 2013. Predator cue
studies reveal strong trait-mediated effects in communities despite variation in experimental
designs. Animal Behaviour 86, 1301-1313 (doi:10.1016/j.anbehav.2013.09.036)
Persons MH, Rypstra AL. 2001. Wolf spiders show graded antipredator behavior in the presence of
chemical cues from different sized predators. Journal of Chemical Ecology 27, 2493-2504
(doi:10.1023/A:1013679532070)
Persons MH, Walker SE, Rypstra AL, Marshall SD. 2001. Wolf spider predator avoidance tactics and
survival in the presence of diet-associated predator cues (Araneae: Lycosidae). Animal
22
Behaviour 61, 43-51 (doi:10.1006/anbe.2000.1594)
Preisser EL, Bolnick DI. 2008. The many faces of fear: comparing the pathways and impacts of
nonconsumptive predator effects on prey populations. PLoS ONE 3, e2465
(doi:10.1371/journal.pone.0002465)
Preisser EL, Orrock JL, Schmitz OJ. 2007. Predator hunting mode and habitat domain alter
nonconsumptive effects in predator-prey interactions. Ecology 88, 2744-2751 (doi:10.1890/070260.1)
Pruitt JN, Stachowicz JJ, Sih A. 2012. Behavioral types of predator and prey jointly determine prey
survival: potential implications for the maintenance of within-species behavioral variation. The
American Naturalist 179, 217-227 (doi:10.1086/663680)
R Development Core Team. 2013. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Relyea RA. 2003. How prey respond to combined predators: a review and an empirical test. Ecology
84, 1827-1839 (doi:10.1890/0012-9658(2003)084[1827:HPRTCP]2.0.CO;2)
Rovner JS. 1996. Conspecific interactions in the lycosid spider Rabidosa rabida: the roles of different
senses. Journal of Arachnology 24, 16-23
Rypstra AL, Schmidt JM, Reif BD, DeVito J, Persons MH. 2007. Tradeoffs involved in site selection
and foraging in a wolf spider: effects of substrate structure and predation risk. Oikos 116, 853863 (doi:10.1111/j.2007.0030-1299.15622.x)
Rypstra AL, Schlosser AM, Sutton PL, Persons MH. 2009. Multimodal signalling: the relative
importance of chemical and visual cues from females to the behaviour of male wolf spiders.
Animal Behaviour 77, 937-947 (doi:10.1016/j.anbehav.2008.12.026)
Schmitt RJ, Holbrook SJ, Brooks AJ, Lape JCP. 2009. Intraguild predation in a structured habitat:
distinguishing multiple-predator effects from competitor effects. Ecology 90, 2434-2443
(doi:10.1890/08-1225.1)
Schmitz OJ. 2010. Resolving ecosystem complexity. Princeton, New Jersey: Princeton University Press.
Schonewolf KW, Bell R, Rypstra AL, Persons MH. 2006. Field evidence of an airborne enemyavoidance kairomone in wolf spiders. Journal of Chemical Ecology 32, 1565-1576
(doi:10.1007/s10886-006-9070-7)
Sih A, Englund G, Wooster D. 1998. Emergent impacts of multiple predators on prey. Trends in
Ecology and Evolution 13, 350-355 (doi:10.1016/S0169-5347(98)01437-2)
23
Steffan SA, Snyder WE. 2010. Cascading diversity effects transmitted exclusively by behavioral
interactions. Ecology 91, 2242-2252 (doi:10.1890/09-0787.1)
Vencl FV, Srygley RB. 2013. Enemy targeting, trade-offs, and the evolutionary assembly of a tortoise
beetle defense arsenal. Evolutionary Ecology 27, 237-252 (doi:10.1007/s10682-012-9603-1)
Walker SE, Marshall SD, Rypstra AL, Taylor DH. 1999. The effects of hunger on locomotory
behaviour in two species of wolf spider (Araneae, Lycosidae). Animal Behaviour 58, 515-520
(doi:10.1006/anbe.1999.1180)
Wesner JS, Billman EJ, Belk MC. 2012. Multiple predators indirectly alter community assembly across
ecological boundaries. Ecology 93, 1647-1682 (doi:10.1890/11-2061.1)
Wrinn KM, Evans SC, Rypstra AL. 2012. Predator cues and an herbicide affect activity and emigration
in an agrobiont wolf spider. Chemosphere 87, 390-396
(doi:10.1016/j.chemosphere.2011.12.030)
24
Table 1. Loading of activities (from Videomex-V software) on principal components. Positive and
negative values indicate positive and negative correlations with the principal component, respectively.
Magnitudes indicate the strength of correlation between the activity variable and the principal
component.
PC1 PC2
Distance (cm)
0.44
0.49
Ambulatory (s)
0.47 -0.33
Stereotypic (s)
0.47 -0.27
Immobile (s)
-0.50 0.33
Speed (cm/s)
0.34
0.68
25
Table 2. Effects of treatment (predator cues present in arena), predator hunger level, and their
interaction on Pardosa activity. Model degrees of freedom: 7, 150.
PC1
df
F
PC2
p
df
F
p
Treatment
3
8.85 <0.01
3
17.36 <0.01
Hunger
1
1.51 0.22
1
89.33 <0.01
Treatment*Hunger
3
1.08 0.36
3
12.86 <0.01
26
Table 3. Effects of treatment (predator cues present in arena) and predator hunger level on Pardosa
activity. Statistics reported are Cohen's d, 95% confidence intervals, and F-values, degrees of freedom,
and p-values from one-way ANOVAs. Treatments are blank (B), cues from Tigrosa (T), and cues from
Scarites (S). Symbols between treatment letters indicate relationships based on effect sizes.
PC1 (low hunger)
d
95% CI
Fdf
PC1 (high hunger)
p
d
95% CI
Fdf
p
T<B
-0.95 (-1.60, -0.30) 5.33,76 <0.01
T<B
-0.95 (-1.64, -0.27) 4.63,74 <0.01
S=B
0.11 (-0.51, 0.73)
S=B
-0.61 (-1.25, 0.02)
T&S < B -0.80 (-1.45, -0.16)
T&S < B -0.95 (-1.59, -0.30)
PC2 (low hunger)
d
95% CI
Fdf
PC2 (high hunger)
p
d
95% CI
Fdf
p
T=B
-0.04 (-0.66, 0.58) 4.53,76 <0.01
T<B
-1.88 (-2.66, -1.11) 22.53,74 <0.01
S=B
0.09 (-0.53, 0.71)
S<B
-2.71 (-3.57, -1.85)
T&S < B -1.01 (-1.67, -0.35)
T&S < B -2.10 (-2.86, -1.33)
27
Table 4. Proportional hazards test of the effects of treatment (cues present in arena), predator hunger
level, and their interaction on Pardosa survival.
Tigrosa as predator
Scarites as predator
df
χ
p
df
χ2
p
Treatment
1
7.28
0.03
1
2.16
0.34
Hunger
2
10.58
<0.01
2
0.00
0.97
Treatment*Hunger
2
0.42
0.81
2
6.80
0.03
2
28
Table 5. Effects of treatment (cues present in arena) and predator hunger level on Pardosa mortality.
Treatments are blank (B), cues from Tigrosa (T), and cues from Scarites (S). Symbols between
treatment letters indicate whether risk was increased, decreased, or unaffected as determined by the
hazard ratio (i.e., instantaneous probability of death).
Tigrosa (low hunger)
Hazard ratio
95% CI
Tigrosa (high hunger)
p
Hazard ratio
95% CI
p
T=B
0.69
(0.36, 1.31) 0.257
T<B
0.56
(0.28, 1.09)
0.086
S=B
0.98
(0.52, 1.83) 0.942
S=B
1.29
(0.68, 2.46)
0.428
T=S
0.71
(0.37, 1.35) 0.291
T<S
0.43
(0.21, 0.87)
0.018
Scarites (low hunger)
Hazard ratio
95% CI
Scarites (high hunger)
p
Hazard ratio
95% CI
p
T=B
0.93
(0.26, 3.33) 0.903
T>B
5.74
(0.92, 110.01) 0.062
S<B
0.37
(0.05, 1.70) 0.206
S>B
8.00
(1.42, 149.79) 0.015
T>S
2.52
(0.54, 17.62) 0.244
T=S
0.72
29
(0.21, 2.25)
0.568
Blank
(B)
Blank
(B)
Tigrosa
(T)
Blank
(B)
Blank
(B)
Blank
(B)
Blank
(B)
Tigrosa
(T)
Blank
(B)
Scarites
(S)
Tigrosa
(T)
Scarites
(S)
Scarites
(S)
Blank
(B)
Scarites
(S)
Tigrosa
(T)
Figure 1. Diagram of arena layouts representing the four treatments used in the patchwork activity
experiment. Filter paper quadrants were blank (B) or previously occupied by Tigrosa (T) or Scarites
(S).
30
Slower <------------->
Faster
Less <-----------------> More
active
active
Figure 2. Box plots of principal components from the patchwork activity experiment for cues from
predators at low and high hunger levels. Treatments are blank (B), cues from Tigrosa (T), Scarites (S),
or both Tigrosa and Scarites (T&S). Box plots show median, first and third quartiles, greatest values
within 1.5 interquartile range, and outliers. Different letters indicate significant differences following
Tukey HSD tests.
31
Figure 3. Pardosa survival in an arena with filter paper that was unmanipulated (blank) or previously
occupied by either Tigrosa or Scarites. The predator with which Pardosa interacted directly is pictured.
Note the differences in time scale between predators. Censored observations are represented by a plus
symbol.
32
Chapter 2: The Importance of Intraguild Predation in Predicting Emergent Multiple
Predator Effects
33
ABSTRACT
Prey typically coexist with multiple predator species, each of which presents a predation risk
related to its habitat domain and foraging mode. These predator characteristics can be used to predict
how the risk from multiple predators will combine to create emergent multiple predator effects for
shared prey. Interactions between predators, particularly intraguild predation, can strongly alter prey
suppression, though the importance of intraguild predation in multiple predator effects has not been
explicitly explored. Furthermore, the vast majority of studies on multiple predator effects has focused
on shared prey that are herbivorous, thus experiments focused on mesopredators are needed to evaluate
the generality of conclusions made about multiple predator effects. I used a suite of carnivorous
arthropods to test a predictive framework of multiple predator effects and to evaluate the role of
intraguild predation in shaping these effects. I allowed the wolf spiders Pardosa milvina, Tigrosa
helluo, and Rabidosa rabida and the ground beetle Scarites quadriceps to interact and recorded the
outcome of all predatory events. I used two tests of multiple predator effects to determine whether
predators created risk enhancement, risk reduction, or were substitutable. I found that the occurrence of
intraguild predation decreased the overall risk to prey, causing observed multiple predator effects to
deviate from predictions. Additionally, I highlight the importance of considering predator identity, as
predators were capable of increasing, decreasing, or not altering the success of their competitors. This
study demonstrates that intraguild predation is a critical factor in determining how the risk from
multiple predators will combine to affect their prey.
34
INTRODUCTION
Studies of predator-prey interactions often focus on interactions between prey and a single
predator species (e.g., Pruitt et al. 2012), and food web studies frequently lump predators into guilds
(e.g., Plass-Johnson et al. 2010) or a single trophic level (e.g., Borer et al. 2006). While these
approaches make systems more tractable and provide mechanistic insight, they do not address the fact
that prey frequently balance demands imposed by a diversity of predators that interact with one another.
To address these issues, researchers have considered how the risks posed by numerous carnivores can
combine to create emergent multiple predator effects on their shared prey (MPEs, Sih et al. 1998). A
growing body of research has demonstrated all possible outcomes of combined predation risk: risk
enhancement (lower survival with two predators than expected by summing survival with each predator
alone), risk reduction (higher survival than expected based on survival with each predator), and
substitutability (lack of risk enhancement or risk reduction) (e.g., Hoverman & Relyea 2009, Woodcock
& Heard 2011, Schneider & Brose 2013). Perhaps more importantly, a theoretical framework for
predicting MPEs has been developed based on simple aspects of carnivore ecology. A review of
numerous multiple-predator studies indicated that the habitat domain (i.e., movement patterns and
microhabitat use) and hunting mode (i.e., strategy used to capture prey) can be used to predict the type
of MPE experienced by prey (Schmitz 2007). Furthermore, these characteristics are useful in predicting
the strength of indirect effects on prey (Preisser et al. 2007) and connectivity of carnivores to disparate
food webs (Wimp et al. 2013). However, the predictive power of habitat domain and hunting mode may
be weakened by the presence of strong interactions between predators in the form of intraguild
predation (Polis et al. 1989, Vance-Chalcraft et al. 2007).
Despite the demonstrated prevalence (Arim & Marquet 2004, Gagnon et al. 2011) and
importance (Law & Rosenheim 2011, Ingram et al. 2012) of intraguild predation, its effects are often
not explicitly evaluated in MPE studies. In fact, trials with intraguild predation (IGP) are often not
analyzed separately (e.g., Woodcock & Heard 2011) or simply discarded (e.g., Soluk & Collins 1998,
O'Gorman et al. 2008); both situations reduce our ability to understand the impact of IGP in studies of
MPEs. In the few experiments conducted to evaluate the impact of intraguild predation on shared prey,
the occurrence of IGP has been shown to reduce predation risk for prey and consequently weaken
trophic cascades (Finke & Denno 2004, 2005). However, to manipulate the presence and absence of
IGP, these studies also simultaneously varied the richness (Finke & Denno 2004) and composition
(Finke & Denno 2005) of predator communities. Predators are not necessarily equivalent (Ramos &
35
Van Buskirk 2012, Schneider & Brose 2013), and if we are to understand the impacts of IGP per se in
multi-predator systems, studies must be conducted without modifying the carnivore community or
changing predator identity (sensu Henry et al. 2010) by altering habitat domain or hunting mode.
The importance of these predator characteristics in shaping MPEs was shown in a study of
numerous generalist arthropod carnivores that consume planthoppers (Woodcock & Heard 2011); the
interaction between habitat domain and hunting mode was a significant predictor of MPEs for prey and
their resource. However, despite the occurrence of IGP, its effects on observed MPEs were not
explicitly quantified. In an experiment that did isolate the impacts of IGP on MPEs, induced changes in
activity and hunting location in response to enhanced habitat structure prevented the separation of IGP
effects from confounding changes in predator characteristics (Finke & Denno 2006). Although both of
these studies contribute to our understanding of multiple predator effects, to be able to isolate the
effects of IGP on predicted MPEs using characteristics of intraguild predators, all aspects of the
carnivore community (including habitat domain and hunting mode) must be held constant.
Here I provide an experimental test of the importance of intraguild predation in predictions of
emergent multiple predator effects based on predator characteristics. Importantly, my study system
consists of multiple carnivores whose shared prey is a predator itself. For ecologists to fully understand
the complexities of MPEs, it is necessary to also study mesopredators, not just herbivores, which have
been used as shared prey in most experiments to date. I use a suite of well-studied intraguild predators
that differ in habitat domain and hunting mode (Figure 1) to test hypotheses about MPEs using a
previously developed predictive framework (Schmitz 2007). Specifically, I expected that predator
characteristics would determine whether the combined risk posed to prey was substitutable between
predators or represented risk enhancement or risk reduction (Figure 1, Table 1). Although the predictive
framework does not explicitly address instances with more than two predators, I anticipated a
combination of interference and IGP would result in risk reduction for the three-predator treatment.
Furthermore, I anticipated the occurrence of intraguild predation would alter MPE outcomes,
corresponding to increased survival of prey in instances where carnivores consumed one another.
Finally, I expected predators to have asymmetric impacts on the foraging success of their competitors,
thus highlighting the importance of predator identity.
36
METHODS
Study species. – The wolf spiders (Araneae: Lycosidae) Pardosa milvina (Hentz), Tigrosa
helluo (Walckenaer), and Rabidosa rabida (Walckenaer) and the ground beetle (Coleoptera: Carabidae)
Scarites quadriceps Chaudoir are generalist arthropod predators that co-occur in and around
agricultural fields of eastern North America (Snyder & Wise 1999, Marshall et al. 2000, Wrinn et al.
2012). Pardosa (all study species hereafter referred to by genus) is substantially smaller than Tigrosa,
Rabidosa, and Scarites (Figure 1), and is the shared prey of these large carnivores (Persons et al. 2001,
Wrinn et al. 2012, this study). The three spiders share the same hunting mode as sit-and-move
predators, though they differ in habitat domain. Tigrosa is largely restricted to the soil surface where
females are facultative burrowers (Walker et al. 1999), Rabidosa is typically found in elevated
vegetation (Brady & McKinley 1994), and Pardosa activity is predominantly at ground level. Scarites
frequently burrows, but emerges at night to actively search for prey on the soil surface (Lundgren et al.
2009, personal observation). Pardosa is most mobile during the day, (Marshall et al. 2002, Schonewolf
et al. 2006), but is likely to encounter Tigrosa, Rabidosa, and Scarites at night, when these predators
are foraging (Lizotte & Rovner 1988, Brady & McKinley 1994, Marshall et al. 2002, Lundgren et al.
2009). Although Pardosa is typically found on the ground, interactions with Tigrosa and Scarites can
induce climbing behavior that may bring Pardosa into contact with Rabidosa (Lowrie 1973, Folz et al.
2006).
Collection and maintenance. – I collected study organisms from corn and soybean fields at
Miami University's Ecology Research Center (39°31′33′′ N, 84°43′20′′W). I maintained spiders and
beetles in an environmental chamber (13h:11h light:dark cycle, 25ºC, 60% RH) for at least one week
prior to use in experiments. I used adult and penultimate female spiders. The sex of Scarites was not
determined, though they were randomly assigned to treatments, so no sex bias is expected.
Furthermore, other researchers have not indicated sex-based differences for Scarites in terms of activity
or ecological interactions (Halaj & Wise 2002, Evans et al. 2010).
Pardosa were housed in plastic containers (7cm wide x 6cm tall), and all other organisms were
housed in larger containers (10.5cm wide x 7.5cm tall). All spider containers had approximately 2cm of
a mixture of moistened potting soil and peat moss (hereafter referred to as soil) as a substrate, and
Scarites containers had approximately 4cm of soil to allow for burrowing. Organisms had water
available ad libitum, and I provided two appropriately-sized (i.e., half the size of the predator) crickets
(Acheta domesticus Linnaeus) once a week, ordered from a local supplier (Chirp N Time, Hamilton,
37
OH, USA). All trials were conducted between July and November 2010.
Multiple predator effects. – I evaluated the separate and combined impacts of the three large
predators (Tigrosa, Rabidosa, and Scarites) on the survival of the smaller carnivore, Pardosa. I added
approximately 2cm of moistened soil to glass terraria (77cm long x 31.5cm wide x 32cm tall) and
provided an opportunity for organisms to climb by placing three evenly spaced bundles of straw (16cm
tall) along one of the walls of each terrarium. Terraria were arranged on metal shelving units and
isolated with opaque barriers to prevent organisms in different terraria from interacting visually. I
emptied and cleaned all terraria with ethanol and allowed them to dry before being used again. Because
soil and straw were re-used between trials, I autoclaved these materials to prevent chemotactile cues
(i.e., silk, feces, and other excreta) deposited by organisms from influencing future trials. In an attempt
to standardize feeding motivation, I provided the large predators two crickets seven days before being
added to the terraria, and all Pardosa were provided two crickets one day before being used. No
crickets were added during experimental trials.
Using an additive design, I created predator treatments by adding a single individual of Tigrosa,
Rabidosa, or Scarites in all possible combinations (n = 12-17 replicates per treatment; 7 treatments),
approximating field densities (Marshall et al. 2000, pers. obsv.). Because the occurrence of IGP is
largely dependent on differences in size between individuals (Balfour et al. 2003, Rypstra & Samu
2005, Wilder & Rypstra 2008), I used size-matched (i.e., based on body mass and length) Tigrosa,
Rabidosa, and Scarites. I allowed these predators to acclimate for 24h before a single Pardosa was
added to each terrarium, and I discarded any trials in which IGP occurred prior to adding Pardosa. I
randomly assigned Pardosa to terraria, and replicates of each treatment were represented in each set of
trials run. Pardosa were released after a 5 min. acclimation under an inverted vial. Every hour for 8h
(four in the light, four in the dark), I checked whether Pardosa had been consumed and if the other
predators had consumed one another. During dark periods, a dim red light was used to minimize
disturbance (Foelix 1996). I concluded the experiment 24h after Pardosa were introduced.
I used a combination of direct observation of predatory events and changes in abdomen width
and body mass to determine whether Tigrosa, Rabidosa, or Scarites consumed Pardosa. Spider
abdomen width and body mass vary with prey consumption (Jakob et al. 1996), thus increases in these
measures over the course of the experiment were indicative of predation. I measured abdomen width
using a digital micrometer (accurate to 0.01mm) attached to a dissecting microscope and body mass
with a balance (accurate to 0.0001g). Scarites abdomen width is fixed at adulthood, so I only measured
38
body mass. I recorded these measurements immediately before adding predators to the terraria and
again within 2h of the end of each trial. No prey were available during this 2h period.
Statistical analyses. – I used two approaches to test hypotheses about how the habitat domain
and hunting mode of my study species contribute to emergent multiple predator effects on their shared
prey (Table 1). I first used logistic regression to determine how the presence of each species impacted
the survival of Pardosa. I then used a test for substitutability developed by Schmitz (2007) which relies
on effect sizes and confidence intervals to evaluate MPEs. Additionally, to determine the importance of
predator identity in multiple predator systems, I examined the impact of each large carnivore on the
foraging success of its competitors.
I tested hypotheses about MPEs using logistic regression models with Pardosa survival as the
response and the presence of each predator as predictors. I used likelihood ratio tests to compare
models assuming substitutability between predators with models devoid of assumptions about their
interactive effects on Pardosa survival (Ramos & Van Buskirk 2012). When these models differed
significantly, the direction of the MPE (i.e., risk enhancement or risk reduction) was determined using
the multiplicative risk model (Soluk & Collins 1988, Sih et al. 1998). When Pardosa survival was
greater or less than expected values from the multiplicative risk model, I concluded risk reduction or
risk enhancement, respectively. I conducted these tests for all pairwise combinations of species, but the
three-way interaction term (i.e., Pardosa survival in the presence of Tigrosa, Rabidosa, and Scarites)
could not be tested due to model saturation.
Additionally, I used a test for substitutability (Schmitz 2007) to test predictions about multiple
predator effects, including the treatment with Tigrosa, Rabidosa, and Scarites. Briefly, I calculated
effect size by dividing the response variable (proportion of trials in which Pardosa survived) in a given
treatment (e.g., Tigrosa with Rabidosa) by the response variable in the control (no predators present). I
corrected effect sizes for predator density by dividing by the number of predators present in each
treatment. The type of MPE was determined using a test for substitutability:
n
n
∑ ( Ri / P i ) = n R1+ 2 .. . +n / ∑ P i
i=1
i= 1
where Ri is the effect size of species i alone, Pi is the density of each species when alone (i.e., always =
1), and R1+2...+n is the effect size for multiple predator treatments. If the values for both sides of the
equation were equal, then the MPE was substitutable. When the values on each side of the equation
differed by more than two standard errors, predators were not considered substitutable.
39
Finally, I examined how each large carnivore impacted the foraging success of its competitors
using logistic regression. For example, I evaluated how the presence of Rabidosa and Scarites impacted
the frequency with which Tigrosa captured Pardosa. These models used the frequency of trials in
which Pardosa was consumed by a given predator as the response and the presence of the other two
predators as predictors. Interaction terms between the other two predators were never significant (p >
0.6 for all interactions) and were therefore excluded from the model. Statistical significance for this test
is considered at the p = 0.05 level, though values between 0.05 and 0.1 are interpreted as biologically
meaningful trends. Criticism of reliance on p-values and strict adherence to the 0.05 cutoff is
articulated elsewhere (e.g., Nakagawa & Cuthill 2007, Stigler 2008, Garamszegi et al. 2009).
While I did not specifically manipulate intraguild predation (here defined as predation between
Tigrosa, Rabidosa, and Scarites), my design allowed me to quantify all predatory events. Therefore, I
were able to evaluate how IGP affected MPEs on Pardosa by conducting analyses with all trials and
with only trials during which IGP did not occur. All analyses were carried out using JMP (version 9.0;
SAS Institute, Inc., Cary, NC, USA) or R (R Core Team 2013).
40
RESULTS
All trials. – Pardosa were consumed in 39% of all trials, and single predators varied in capture
success from 56% (Tigrosa) to 6% (Scarites) (Figure 2). When more than one predator was present,
Pardosa were consumed 47% of the time. Of all the treatments with multiple predators, Pardosa were
consumed most often in the treatment where Tigrosa and Rabidosa were present and least often when
Tigrosa was absent (Figure 2a).
My two statistical approaches to understand how habitat domain and hunting mode contribute to
multiple predator effects produced contrasting results. The logistic regression models used to test
hypotheses regarding MPEs revealed both risk reduction and substitutability between pairs of predators
(Table 1, Table 2). This approach matched predictions based on species characteristics for all pairwise
combinations except Tigrosa and Rabidosa, which exhibited risk reduction where substitutability was
expected. Using the test for substitutability to test the same hypotheses about MPEs revealed
substitutability between all pairwise predator combinations, and risk reduction in the three predator
treatment (Table 1). These results matched predictions based on habitat domain and hunting mode
except for evidence of substitutability between Tigrosa and Scarites where risk reduction was
predicted.
Examining the impact of each large carnivore on the foraging success of its competitors
indicated that the presence of Rabidosa and Scarites tended to decrease the frequency with which
Tigrosa consumed Pardosa (Table 3, Figure 2b). Conversely, the consumption of Pardosa by Rabidosa
and Scarites was largely unaffected by the presence of other predators (Table 3, Figure 2c, d).
Excluding trials with intraguild predation. – Intraguild predation occurred in 21% of all trials
with multiple predators. Tigrosa and Rabidosa engaged in reciprocal intraguild predation: Tigrosa
consumed Rabidosa in 2/14 trials in the two-predator treatment and in 4/16 trials in the three-predator
treatment, and Rabidosa consumed Tigrosa in 2/14 trials in the two-predator treatment and in 2/16
trials in the three-predator treatment. Scarites only consumed Rabidosa (2/14 trials in the two-predator
treatment, 1/16 trials in the three-predator treatment), and was never consumed by either spider.
Removing trials with IGP revealed a 9% decrease in Pardosa survival. Pardosa were consumed most
often in treatments with all three predators, and survival was highest when only Rabidosa and Scarites
were present (Figure 3).
As I found in the analysis of all trials, my statistical tests of MPEs also led to different
conclusions for the restricted data set. When I used logistic regression to test hypotheses about
41
emergent multiple predator effects, the observed MPE only matched the expected outcome (i.e.,
substitutability) for the treatment with Tigrosa and Scarites; whereas risk enhancement was evident in
the other two-predator treatments (Table 1, Table 2). The overall decrease in Pardosa survival evident
when trials with IGP were excluded was also reflected in conclusions from the test for substitutability:
the treatment with Rabidosa and Scarites as well as the three-predator treatment showed increased risk
compared to when all trials were included (Table 1). However, when Tigrosa was paired with either
Rabidosa or Scarites, MPEs were unaffected by the removal of trials with IGP (Table 1).
The analysis used to examine interactions between the large carnivores revealed that, in trials
without IGP, Tigrosa tended to consume Pardosa less often in the presence of other predators, though
the impact of Rabidosa was not strong (Table 3, Figure 3b). In contrast, Rabidosa tended to be more
successful when Tigrosa and Scarites were present (Table 3, Figure 3c). Consumption of Pardosa by
Scarites was not affected by the other predators (Table 3, Figure 3d).
42
DISCUSSION
Here I demonstrate the importance of considering intraguild predation when using predator
characteristics (i.e., habitat domain and hunting mode) to predict emergent multiple predator effects.
Predicted MPEs were observed in many cases, following the framework proposed by Schmitz (2007)
(Figure 1, Table 1). However, conclusions about the types of MPE were often modified by the
exclusion of trials in which intraguild predation occurred, thus demonstrating the importance of IGP in
multiple-predator systems. Additionally, the more conservative analysis frequently described
substitutability between carnivores, whereas logistic regression models often revealed risk reduction
and risk enhancement. Predator identity was also a key factor in determining predation risk to shared
prey. Species were shown to increase, decrease, or have no effect on the predation success of their
competitors. This is the first study that I know of to use predator characteristics to predict MPEs with a
design that isolates the impacts of intraguild predation and predator identity.
All trials. – The analysis of emergent multiple predator effects using the test for substitutability
largely agreed with a priori predictions based on habitat domain and hunting mode. The
complementarity of habitat domains of Tigrosa and Rabidosa (Figure 1) likely explained the
substitutable effects they had on Pardosa survival. Because Pardosa can freely move between predator
domains, it can average the risk posed by these species. Alternatively, Pardosa may use compensatory
defenses (e.g., reduced activity) to minimize mortality from predators differing in the level of risk they
pose and in habitat domain (Krupa & Sih 1998). In trials with Tigrosa and Scarites, their overlapping
habitat domains and different hunting modes (Figure 1) were predicted to translate to a reduced risk of
predation for their shared prey. This risk reduction is predicted to arise when prey escape both
predators, which engage in IGP, or when predators simply interfere with one another (Vance-Chalcraft
& Soluk 2005, Schmitz 2007, Carey & Wahl 2010); however, Tigrosa and Scarites never consumed one
another. Substitutability between Tigrosa and Scarites may result from a balance between Scarites
activity increasing Pardosa movement, thus placing it at greater risk from Tigrosa (Persons et al. 2001),
while simultaneously interfering with Tigrosa. The substitutable effects of Rabidosa and Scarites
followed from their complementary habitat domains (Figure 1), and my prediction for risk reduction in
the three-predator treatment was likely supported by a combination of interference and IGP between
the predators.
The test for substitutability has been criticized for being overly conservative and often reporting
substitutability (Tylianakis & Romo 2010), so I also used logistic regression to test whether MPEs can
43
be predicted from species characteristics. Overall, the test for substitutability indicated that predators
were substitutable 75% of the time, regardless of whether or not trials with IGP were incuded (Table 1).
Substitutability was only found 17% of the time using logistic regression (Table 2), and the MPEs from
this analysis differed from those predicted by habitat domain and hunting mode only in the treatment
with Tigrosa and Rabidosa. These predators engaged in reciprocal IGP, despite their differences in
habitat domain (Figure 1). In the field, Tigrosa can occasionally be found on elevated vegetation, and
Rabidosa travel along the ground infrequently (Brady & McKinley 1994, personal observation), so
interactions between these predators are unlikely to be an artifact of the laboratory setting. Risk
reduction was likely detected with this approach because it is less conservative than the test for
substitutability.
Using both logistic regression and the test for substitutability enabled a more complete
understanding of MPEs. The test for substitutability, although conservative, allows for examination of
MPEs in the three-predator treatment and fits with the predictive framework and language established
by Schmitz (2007). Additionally, it is part of a developing trend of using effect sizes and confidence
intervals to interpret results (Nakagawa & Cuthill 2007, Garamszegi et al. 2009). In this study,
traditional hypothesis testing yielded less conservative estimates of MPEs, and researchers have begun
using this approach in conjunction with effect size-based analyses to more fully understand biological
interactions (Wesner et al. 2012).
Excluding trials with intraguild predation. – I examined trials without IGP to understand its role
in MPEs, even though these predators engage in IGP frequently in nature. The occurrence of IGP
affected conclusions about MPEs, regardless of which analytic technique was used. For the majority of
the treatments, the risk of predation for Pardosa was increased by limiting the analyses to trials in
which IGP did not occur. Evidence of increased risk in the presence of Rabidosa and Scarites may be
an example of predator facilitation (e.g., Meyer & Byers 2005, Steinmetz et al. 2008), wherein Pardosa
evading surface-active Scarites are at increased risk of consumption by Rabidosa (e.g., from climbing).
The distinct habitat domains of Rabidosa and Scarites (Figure 1) likely prevented or minimized
interference between these predators, further contributing to risk enhancement for Pardosa.
Additionally, a meta-analysis of the role of IGP in prey suppression by Vance-Chalcraft et al. (2007)
revealed that mutual IGP, but not unidirectional IGP, causes a significant reduction in consumption of
prey. Indeed, mutual IGP between Tigrosa and Rabidosa reduced the risk of predation on Pardosa,
though this was also observed in predator pairs with unidirectional IGP (i.e., Rabidosa and Scarites).
44
Excluding trials with IGP did not alter MPEs for trials with Tigrosa and Scarites as predators. This pair
of carnivores did not engage in IGP, thus the conclusions about their joint effects on Pardosa survival
were unchanged. The only other instance in which IGP did not alter the type of MPE was from the test
for substitutability for trials with Tigrosa and Rabidosa. However, the logistic regression indicated risk
enhancement when IGP was excluded for this treatment, further demonstrating the conservative nature
of the first analysis (Tylianakis & Romo 2010).
The decreased risk of predation on shared prey is likely driven by a suite of changes associated
with an IGP event. First, the diversity of predators is decreased, thus prey may gain spatial refuge from
the remaining carnivores by moving into habitats previously occupied by the consumed predator.
Second, there are fewer total predators for prey to avoid. Third, predators surviving an IGP event will
be at a lower hunger level and therefore less likely to pursue and consume their prey. These
mechanisms are not mutually exclusive and may singly or jointly explain the observed decrease in
Pardosa survival when IGP trials were removed from the analyses.
Similar effects of IGP have been demonstrated in carnivore-herbivore systems, with prey
experiencing a release of predation pressure when IGP occurs (Finke & Denno 2004, 2005; Barton &
Schmitz 2009; Sanders et al. 2011). Interestingly, these studies differ in conclusions about how the
effects of IGP cascade through prey populations to affect their resources (e.g., plant biomass): cascades
have been shown to be strengthened (Barton & Schmitz 2009) and weakened (Finke & Denno 2004,
2005; Sanders et al. 2011) by the occurrence of predation between predators. Therefore, it will be
worthwhile to continue investigating the impact of IGP on trophic cascades. Furthermore, studies are
needed where the shared prey species is itself a predator (i.e., a mesopredator) and in non-grazing (e.g.,
detrital) systems, especially since predators can be more tightly linked to detrital food webs than
grazing food webs (Chen & Wise 1999, Oelbermann et al. 2008, Sanders et al. 2011). It is essential to
understand the role of IGP in modifying multiple predator effects to predict how these interactions may
alter food webs and ecosystem function.
Predator identity. – One advantage of the design of this experiment is the ability to attribute
prey consumption to specific predators in multiple-predator treatments, a “difficult, if not impossible”
task important to understanding the impacts of carnivore diversity on communities (Finke & Snyder
2010). Obtaining this high resolution information allows evaluation of how the presence of
interspecific competitors affects the ability of predators to capture prey. Tigrosa capture success tended
to be negatively affected by the other predators (note trend line in Figure 2b), though this was not
45
simply a result of IGP (note trend line in Figure 3b). Behavioral interactions (e.g., agonistic
interactions), exploitative interference, and strengthened anti-predator behavior by prey are all likely
mechanisms explaining the decreased frequency of Pardosa being consumed by Tigrosa in the presence
of Rabidosa and Scarites. The sit-and-move hunting mode of Tigrosa may make it particularly
vulnerable to disruption from other predators, as frequent relocation due to interactions with active
predators (e.g., Scarites) may reduce opportunities for successful ambush of prey. Despite the obvious
impacts of IGP on the likelihood of Tigrosa consuming Pardosa, non-consumptive effects have been
shown to be equally or more important than consumptive effects in predator-prey interactions (Preisser
et al. 2005, Steffan & Snyder 2010). Future studies on the impact of IGP on prey suppression would
benefit by separating consumptive and non-consumptive effects, as behavioral interactions or predator
cues alone can strongly impact multiple trophic levels (e.g., Steffan & Snyder 2010, Rypstra & Buddle
2013).
In contrast to the effect of competitors on Tigrosa, Rabidosa capture success tended to be
positively affected by the presence of Tigrosa and Scarites (note trend line in Figure 2c), evidence for
predator facilitation (e.g., Cresswell & Quinn 2013); though this increase was more pronounced when
IGP trials were excluded (note trend line in Figure 3c). The spatial separation between Rabidosa and
both Tigrosa and Scarites (Figure 1) likely explains the increased capture success; as disruption to the
sit-and-move hunting mode of elevated Rabidosa was unlikely to occur. Facilitation between Scarites
and Rabidosa may drive the change in MPE from substitutable to risk enhancement, as suggested by
both statistical approaches. The fact that facilitation was only evident when excluding trials with IGP
indicates that the consumptive component of interactions with competitors may be a more important
determinant of Rabidosa success than non-consumptive interactions. Although Scarites rarely
consumed Pardosa, its presence was still important in shaping MPEs. The impacts of Scarites are most
likely driven by behavioral interactions, as IGP frequency was low and its activity has the potential to
interfere with Tigrosa and increase activity of Pardosa. Other studies have demonstrated how
carnivores that infrequently consume prey can still impact MPEs by interacting with other predators or
having non-consumptive effects on prey (e.g., Van Son & Thiel 2006, Schmitt et al. 2009). Overall,
gathering data to record the identity of the predator in each interaction allowed me to recognize
asymmetric impacts predators had on one another. Specifically, Scarites success was unaffected by the
presence of the other large carnivores, but Scarites had negative and positive effects on the success of
Tigrosa and Rabidosa, respectively. In nature, where Pardosa will regularly encounter multiple
46
predators, the reduction in success of the most dangerous predator, Tigrosa, coupled with the increase
in success of Rabidosa, a species occupying an elevated microhabitat, may result in Pardosa spending
more time at the soil surface. Thus, there is potential for multiple-predator effects to alter the
connection strength between mesopredators and different food webs, a concept in need of further study.
Conclusions. – I have shown how simple characteristics of predators can often be used to
predict their combined effects on shared prey and how intraguild predation can alter these emergent
multiple predator effects. IGP reduces the risk posed to shared prey, and thus the extent of IGP in a
system is likely predictive of the potential for prey suppression. Efforts to manipulate or manage
carnivore diversity as a means of pest suppression should consider non-consumptive interactions
between predators as well as the asymmetric, idiosyncratic impacts predators can have on one another.
The vast majority of research on MPEs has used herbivores as prey, so studies such as this one focusing
on shared prey that are predators themselves are needed to gain a more complete understanding of
MPEs and determine the extent to which these patterns can be generalized. Appreciating the influence
of predator diversity and identity on prey survival is key to understanding how communities will
respond to changes in carnivore diversity (Griffin et al. 2013). Because increases in food web
complexity have been shown modify to emergent multiple predator effects and trophic cascades
(Philpott et al. 2012), future research on these topics will benefit from increasing food web realism by
including greater vertical and horizontal diversity.
ACKNOWLEDGEMENTS
I am grateful to many graduate and undergraduate members of my research group for assistance
with animal care and manuscript feedback. Statistical advice was provided by Thomas Crist and
Michael Hughes. Douglas Sitvarin illustrated the study species. Funding was provided by the
Department of Zoology and Hamilton Campus of Miami University and a Sigma Xi G.I.A.R..
47
REFERENCES
Arim M, Marquet PA. 2004. Intraguild predation: a widespread interaction related to species biology.
Ecology Letters 7, 557-564 (doi:10.1111/j.1461-0248.2004.00613.x)
Balfour RA, Buddle CM, Rypstra AL, Walker SE, Marshall SD. 2003. Ontogenetic shifts in
competitive interactions and intra-guild predation between two wolf spider species. Ecological
Entomology 28, 25-30 (doi:10.1046/j.1365-2311.2002.00486.x)
Barton BT, Schmitz OJ. 2009. Experimental warming transforms multiple predator effects in a
grassland food web. Ecology Letters 12, 1317-1325 (doi:10.1111/j.1461-0248.2009.01386.x)
Borer ET, Halpern BS, Seabloom EW. 2006. Asymmetry in community regulation: effects of predators
and productivity. Ecology 87, 2813-2820 (doi:10.1890/00129658(2006)87[2813:AICREO]2.0.CO;2)
Brady AR, McKinley KS. 1994. Nearctic species of the wolf spider genus Rabidosa (Araneae,
Lycosidae). Journal of Arachnology 22, 138-160
Carey MP, Wahl DH. 2010. Interactions of multiple predators with different foraging modes in an
aquatic food web. Oecologia 162, 443-452 (doi:10.1007/s00442-009-1461-3)
Chen B, Wise DH. 1999. Bottom-up limitation of predaceous arthropods in a detritus-based terrestrial
food web. Ecology 80, 761-772 (doi:10.1890/0012-9658(1999)080[0761:BULOPA]2.0.CO;2)
Cresswell W, Quinn JL. 2013. Contrasting risks from different predators change the overall nonlethal
effects on predation risk. Behavioral Ecology 24, 871-876 (doi:10.1093/beheco/art023)
Evans SC, Shaw EM, Rypstra AL. 2010. Exposure to a glyphosate-based herbicide affects agrobiont
predatory arthropod behaviour and long-term survival. Ecotoxicology 19, 1249-1257
(doi:10.1007/s10646-010-0509-9)
Finke DL, Denno RF. 2004. Predator diversity dampens trophic cascades. Nature 429, 407-410
(doi:10.1038/nature02554)
Finke DL, Denno RF. 2005. Predator diversity and the functioning of ecosystems: the role of intraguild
predation in dampening trophic cascades. Ecology Letters 8, 1299-1306 (doi:10.1111/j.14610248.2005.00832.x)
Finke DL, Denno RF. 2006. Spatial refuge from intraguild predation: implications for prey suppression
and trophic cascades. Oecologia 149, 265-275 (doi:10.1007/s00442-006-0443-y)
Finke DL, Snyder WE. 2010. Conserving the benefits of predator biodiversity. Biological Conservation
143, 2260-2269 (doi:10.1016/j.biocon.2010.03.022)
48
Foelix RF. 1996. Biology of Spiders, 2nd edn. Oxford University Press, Oxford.
Folz HC, Wilder SM, Persons MH, Rypstra AL. 2006. Effects of predation risk on vertical habitat use
and foraging of Pardosa milvina. Ethology 112, 1152-1158 (doi:10.1111/j.14390310.2006.01276.x)
Gagnon A-E, Heimpel GE, Brodeur J. 2011. The ubiquity of intraguild predation among predatory
arthropods. PLoS ONE 6, e28061 (doi:10.1371/journal.pone.0028061)
Garamszegi LZ, Calhim S, Dochtermann N, Hegyi G, Hurd PL, Jorgensen C, Kutsukake N, Lajeunesse
MJ, Pollard KA, Schielzeth H, Symonds MRE, Nakagawa S. 2009. Changing philosophies and
tools for statistical inferences in behavioral ecology. Behavioral Ecology 20, 1363-1375
(doi:10.1093/beheco/arp137)
Griffin JN, Byrnes JEK, Cardinale BJ. 2013. Effects of predator richness on prey suppression: a metaanalysis. Ecology 94, 2180-2187 (doi:10.1890/13-0179.1)
Halaj T, Wise DH. 2002. Impact of a detrital subsidy on trophic cascades in a terrestrial grazing food
web. Ecology 83, 3141-3151 (doi:10.2307/3071849)
Henry LM, Bannerman JA, Gillespie DR, Roitberg BD. 2010. Predator identity and the nature and
strength of food web interactions. Journal of Animal Ecology 79, 1164-1171
(doi:10.1111/j.1365-2656.2010.01723.x)
Hoverman JT, Relyea RA. 2009. Survival trade-offs associated with inducible defences in snails: the
roles of multiple predators and developmental plasticity. Functional Ecology 23, 1179-1188
(doi:10.1111/j.1365-2435.2009.01586.x)
Ingram T, Svanback R, Kraft NJB, Kratina P, Southcott L, Schluter D. 2012. Intraguild predation drives
evolutionary niche shift in threespine stickleback. Evolution 66, 1819-1832 (doi:10.1111/j.15585646.2011.01545.x)
Jakob EM, Marshall SD, Uetz GW. 1996. Estimating fitness: a comparison of body condition indices.
Oikos 77, 61-67 (doi:10.2307/3545585)
Krupa JJ, Sih A. 1998. Fishing spiders, green sunfish, and a stream-dwelling water strider: malefemale conflict and prey responses to single versus multiple predator environments. Oecologia
117, 258-265 (doi:10.1007/s004420050656)
Law Y-H, Rosenheim JA. 2011. Effects of combining an intraguild predator with a cannibalistic
intermediate predator on a species-level trophic cascade. Ecology 92, 333-341 (doi:10.1890/100156.1)
49
Lizotte RS, Rovner JS. 1988. Nocturnal capture of fireflies by lycosid spiders – visual versus vibratory
stimuli. Animal Behaviour 36, 1809-1815 (doi:10.1016/S0003-3472(88)80120-9)
Lowrie DC. 1973. The microhabitats of western wolf spiders of the genus Pardosa. Entomological
News 84, 103-116
Lundgren JG, Nichols S, Prischmann DA, Ellsbury MM. 2009. Seasonal and diel activity patterns of
generalist predators associated with Diabrotica vigifera immatures (Coleoptera:
Chrysomelidae). Biocontrol Science and Technology 19, 327-333
(doi:10.1080/09583150802696533)
Marshall SD, Walker SE, Rypstra AL. 2000. A test for a differential colonization and competitive
ability in two generalist predators. Ecology 81, 3341-3349
Marshall SD, Pavuk DM, Rypstra AL. 2002. A comparative study of the phenology and daily activity
patterns in the wolf spiders Pardosa milvina and Hogna helluo in soybean agroecosystems in
southwestern Ohio (Araneae, Lycosidae). Journal of Arachnology 30, 503-510
(doi:10.1636/0161-8202(2002)030[0503:ACSOPA]2.0.CO;2)
Meyer JJ, Byers JE. 2005. As good as dead? Sublethal predation facilitates lethal predation on an
intertidal clam. Ecology Letters 8, 160-166 (doi:10.1111/j.1461-0248.2004.00700.x)
Nakagawa S, Cuthill IC. 2007. Effect size, confidence interval and statistical significance: a
practical guide for biologists. Biological Reviews 82, 591-605 (doi:10.1111/j.1469185X.2007.00027.x)
O'Gorman EJ, Enright RA, Emmerson MC. 2008. Predator diversity enhances secondary production
and decreases the likelihood of trophic cascades. Oecologia 158, 557-567 (doi:10.1007/s00442008-1165-0)
Oelbermann K, Langel R, Scheu S. 2008. Utilization of prey from the decomposer system by generalist
predators of grassland. Oecologia 155, 605-617 (doi:10.1007/s00442-007-0927-4)
Persons MH, Walker SE, Rypstra AL, Marshall SD. 2001. Wolf spider predator avoidance tactics and
survival in the presence of diet-associated predator cues (Araneae: Lycosidae). Animal
Behaviour 61, 43-51 (doi:10.1006/anbe.2000.1594)
Philpott SM, Pardee GL, Gonthier DJ. 2012. Cryptic biodiversity effects: importance of functional
redundancy revealed through addition of food web complexity. Ecology 93, 992-1001
Plass-Johnson JG, McQuaid CD, Porri F. 2010. Top-down effects on intertidal mussel populations:
assessing two predator guilds in a South African marine protected area. Marine Ecology
50
Progress Series 411, 149-159 (doi:10.3354/meps08681)
Polis GA, Myers CA, Holt RD. 1989. The ecology and evolution of intraguild predation: potential
competitors that eat each other. Annual Review of Ecology and Systematics 20, 297-330
(doi:10.1146/annurev.es.20.110189.001501)
Preisser EL, Bolnick DI, Benard ME. 2005. Scared to death? The effects of intimidation and
consumption in predator-prey interactions. Ecology 86, 501-509 (doi:10.1890/04-0719)
Preisser EL, Orrock JL, Schmitz OJ. 2007. Predator hunting mode and habitat domain alter
nonconsumptive effects in predator-prey interactions. Ecology 88, 2744-2751 (doi:10.1890/070260.1)
Pruitt JN, Stachowicz JJ, Sih A. 2012. Behavioral types of predator and prey jointly determine prey
survival: potential implications for the maintenance of within-species behavioral variation. The
American Naturalist 179, 217-227 (doi:10.1086/663680)
R Development Core Team. 2013. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Ramos O, Van Buskirk J. 2012. Non-interactive multiple predator effects on tadpole survival.
Oecologia 169, 535-539 (doi:10.1007/s00442-011-2208-5)
Rypstra AL, Buddle CM. 2013. Spider silk reduces insect herbivory. Biology Letters 9, 20120948
(doi:10.1098/rsbl.2012.0948)
Rypstra AL, Samu F. 2005. Size dependent intraguild predation and cannibalism in coexisting wolf
spiders (Araneae, Lycosidae). Journal of Arachnology 33, 390-397 (doi:10.1636/CT05-10.1)
Sanders D, Schaefer M, Platner C, Griffiths GJK. 2011. Intraguild interactions among generalist
predator functional groups drive impact on herbivore and decomposer prey. Oikos 120, 418-426
(doi:10.1111/j.1600-0706.2010.18924.x)
Schmitt RJ, Holbrook SJ, Brooks AJ, Lape JCP. 2009. Intraguild predation in a structured habitat:
distinguishing multiple-predator effects from competitor effects. Ecology 90, 2434-2443
(doi:10.1890/08-1225.1)
Schmitz OJ. 2007. Predator diversity and trophic interactions. Ecology 88, 2415-2426 (doi:10.1890/060937.1)
Schneider FD, Brose U. 2013. Beyond diversity: how nested predator effects control ecosystem
functions. Journal of Animal Ecology 82, 64-71 (doi:10.1111/1365-2656.12010)
Schonewolf KW, Bell R, Rypstra AL, Persons MH. 2006. Field evidence of an airborne enemy-
51
avoidance kairomone in wolf spiders. Journal of Chemical Ecology 32, 1565-1576
(doi:10.1007/s10886-006-9070-7)
Sih A, Englund G, Wooster D. 1998. Emergent impacts of multiple predators on prey. Trends in
Ecology and Evolution 13, 350-355 (doi:10.1016/S0169-5347(98)01437-2)
Snyder WE, Wise DH. 1999. Predator interference and the establishment of generalist predator
populations for biocontrol. Biological Control 15, 283-292 (doi:10.1006/bcon.1999.0723)
Soluk DA, Collins NC. 1988. Synergistic interactions between fish and stoneflies: facilitation and
interference among stream predators. Oikos 52, 94-100 (doi:10.2307/3565987)
Steffan SA, Snyder WE. 2010. Cascading diversity effects transmitted exclusively by behavioral
interactions. Ecology 91, 2242-2252 (doi:10.1890/09-0787.1)
Steinmetz J, Soluk DA, Kohler SL. 2008. Facilitation between herons and smallmouth bass foraging on
common prey. Environmental Biology of Fishes 81, 51-61 (doi:10.1007/s10641-006-9165-6)
Stigler, S. 2008. Fisher and the 5% level. Chance 21, 12 (doi:10.1080/09332480.2008.10722926)
Tylianakis JM, Romo CM. 2010. Natural enemy diversity and biological control: making sense of the
context-dependency. Basic and Applied Ecology 11, 657-668 (doi:10.1016/j.baae.2010.08.005)
Van Son TC, Thiel M. 2006. Multiple predator effects in an intertidal food web. Journal of Animal
Ecology 75, 25-32 (doi:10.1111/j.1365-2656.2005.01019.x)
Vance-Chalcraft HD, Soluk DA. 2005. Estimating the prevalence and strength of non-independent
predator effects. Oecologia 146, 452-460 (doi:10.1007/s00442-005-0201-6)
Vance-Chalcraft HD, Rosenheim JA, Vonesh JR, Osenberg CW, Sih A. 2007. The influence of
intraguild predation on prey suppression and prey release: a meta-analysis. Ecology 88, 26892696 (doi:10.1890/06-1869.1)
Walker SE, Marshall SD, RypstraAL. 1999. The effect of feeding history on retreat construction in the
wolf spider Hogna helluo (Araneae, Lycosidae). Journal of Arachnology 27, 689-691
Wesner JS, Billman EJ, Belk MC. 2012. Multiple predators indirectly alter community assembly across
ecological boundaries. Ecology 93, 1647-1682
Wilder SM, Rypstra AL. 2008. Sexual size dimorphism mediates the occurrence of state-dependent
sexual cannibalism in a wolf spider. Animal Behaviour 76, 447-454
(doi:10.1016/j.anbehav.2007.12.023)
Wimp GM, Murphy SM, Lewis D, Douglas MR, Ambikapathi R, Van-Tull L'A, Gratton C, Denno RF.
2013. Predator hunting mode influences patterns of prey use from grazing and epigeic food
52
webs. Oecologia 171, 505-515 (doi:10.1007/s00442-012-2435-4)
Woodcock BA, Heard MS. 2011. Disentangling the effects of predator hunting mode and habitat
domain on the top-down control of insect herbivores. Journal of Animal Ecology 80, 495-503
(doi:10.1111/j.1365-2656.2010.01790.x)
Wrinn KM, Evans SC, Rypstra AL. 2012. Predator cues and an herbicide affect activity and emigration
in an agrobiont wolf spider. Chemosphere 87, 390-396
(doi:10.1016/j.chemosphere.2011.12.030)
53
Table 1. Expected and observed multiple predator effects (MPEs) on Pardosa survival due to predation
by Tigrosa (T), Rabidosa (R), and Scarites (S). Tests for MPEs were also analyzed by excluding trials
with intraguild predation (IGP: here defined as predation between Tigrosa, Rabidosa, and Scarites).
Predators
Expected MPE
Observed MPE Observed MPE
(all trials)
(excluding IGP)
Test for substitutability
T&R
Substitutable
Substitutable
Substitutable
T&S
Risk reduction
Substitutable
Substitutable
R&S
Substitutable
Substitutable
Risk enhancement
Risk reduction
Risk reduction
Substitutable
T&R&S
Logistic regression
T&R
Substitutable
Risk reduction Risk enhancement
T&S
Risk reduction
Risk reduction
Risk reduction
R&S
Substitutable
Substitutable
Risk enhancement
54
Table 2. Test for multiple predator effects (MPEs) on Pardosa survival due to predation by Tigrosa (T),
Rabidosa (R), and Scarites (S) using logistic regression. Analyses were also conducted by excluding
trials with intraguild predation (IGP: here defined as predation between Tigrosa, Rabidosa, and
Scarites).
χ2
d.f.
p-value
T&R
9.54
1
0.002
T&S
4.10
1
0.043
R&S
0.24
1
0.623
T&R
6.64
1
0.010
T&S
8.49
1
0.004
R&S
3.96
1
0.047
Predators
All trials
Excluding IGP
55
Table 3. Impact of Tigrosa (T), Rabidosa (R), and Scarites (S) on the frequency with which another
predator consumed Pardosa. Values reported are regression coefficients (p-values). Regression
coefficients represent the change in the response (e.g., frequency of Tigrosa consuming Pardosa) given
one unit change in a predictor (e.g., presence of Rabidosa). Analyses were also conducted by excluding
trials with intraguild predation (IGP: here defined as predation between Tigrosa, Rabidosa, and
Scarites).
Impact of T
All trials
Excluding IGP
Consumed by T
Consumed by R
0.370 (0.551)
Consumed by S
0.000 (1.000)
Consumed by T
Impact of R
Impact of S
-1.128 (0.058)
-1.091 (0.061)
0.746 (0.237)
-17.790 (0.996)
-0.844 (0.201)
Consumed by R
1.114 (0.109)
Consumed by S
0.000 (1.000)
56
-1.277 (0.041)
1.238 (0.078)
-17.790 (0.996)
Scarites quadriceps
Domain: narrow (soil
surface & subterranean)
Mode: active
Rabidosa rabida
Domain: narrow
(elevated vegetation)
Mode: sit-and-move
Tigrosa helluo
Domain: narrow
(soil surface)
Mode: sit-andmove
Pardosa milvina
Domain: broad
Mode: sit-andmove
Figure 1. Habitat domain and hunting mode of study species. Arrows point from predator to prey. The
inset figure depicts the spatial relationships of Pardosa (rectangle) and Tigrosa, Rabidosa, and Scarites
(ovals). The lack of domain overlap between Tigrosa and Rabidosa is predicted to lead to substitutable
risk. The combination of overlapping domains and different hunting modes for Tigrosa and Scarites is
predicted to create risk reduction by means of intraguild predation. Rabidosa and Scarites have nonoverlapping domains and should create substitutable risk.
57
1
A) Total
Proportion consumed
Proportion consumed
1
0.8
0.6
0.4
0.2
0.6
0.4
0
Single predator
T& R
T& S
R&S
T& R& S
Single predator
1
T& R
T& S
R&S
T& R& S
T& S
R&S
T& R& S
1
C) Rabidosa
Proportion consumed
Proportion consumed
B) Tigrosa
0.2
0
0.8
0.8
0.6
0.4
0.8
D) Scarites
0.6
0.4
0.2
0.2
0
0
Single predator
T& R
T& S
R&S
T& R& S
Single predator
T& R
Figure 2. Proportion of trials in which Pardosa was consumed, summed across predators (A) and per
predator (B-D). Including trials with predation between Tigrosa (T), Rabidosa (R), and Scarites (S).
Fitted lines illustrate trends in predation success (see discussion).
58
1
A) Total
Proportion consumed
Proportion consumed
1
0.8
0.6
0.4
0.2
0
T& R
T& S
R&S
0.6
0.4
0.2
T& R& S
Single predator
1
T& R
T& S
R&S
T& R& S
T& S
R&S
T& R& S
1
C) Rabidosa
Proportion consumed
Proportion consumed
B) Tigrosa
0
Single predator
0.8
0.8
0.6
0.4
0.2
0.8
D) Scarites
0.6
0.4
0.2
0
0
Single predator
T& R
T& S
R&S
T& R& S
Single predator
T& R
Figure 3. Proportion of trials in which Pardosa was consumed, summed across predators (A) and per
predator (B-D). Excluding trials with predation between Tigrosa (T), Rabidosa (R), and Scarites (S).
Fitted lines illustrate trends in predation success (see discussion).
59
Chapter 3: Fear of Predation Alters Soil CO2 Flux and Nitrogen Content
60
ABSTRACT
Predators are known to have both consumptive and nonconsumptive effects on their prey that
can cascade to affect lower trophic levels. Nonconsumptive interactions often drive these effects,
though the majority of studies have been conducted in aquatic or herbivory-based systems. Here I use a
laboratory study to examine how linkages between an above-ground predator and a detritivore
influence below-ground properties. I demonstrate that predators can depress soil metabolism (i.e., CO2
flux) and soil nutrient content via both consumptive and nonconsumptive interactions with detritivores,
and that the strength of isolated nonconsumptive effects is comparable to changes resulting from
predation. Changes in detritivore abundance and activity in response to predators and the fear of
predation likely mediate interactions with the soil microbe community. My results underscore the need
to explore these mechanisms at large scales, considering the disproportionate extinction risk faced by
predators and the importance of soils in the global carbon cycle.
61
INTRODUCTION
Predators can affect prey populations directly by consuming individuals and indirectly by
causing changes in prey traits (e.g., behavior) as prey exhibit a fear response to the risk of predation
(Schmitz 2010). These interactions between predators and prey have been termed consumptive and
nonconsumptive effects, respectively. Surprisingly, nonconsumptive effects (NCEs) often have an equal
or greater magnitude than consumptive effects (CEs) on both prey and prey resources (Preisser et al.
2005), and the importance of NCEs has been widely demonstrated (Werner & Peacor 2003).
The vast majority of research into predator effects on prey and their resources has focused on
the “green pathway” that links predators to plants via herbivores. In contrast, the “brown pathway”
linking predators to detrital pools via detritivores has received considerably less attention (Schmitz
2010), despite the applicability of “green” theory to “brown” systems (Hassall et al. 2006) and the
importance of soils in the global carbon cycle (Allison 2006). Although predation studies in detrital
systems are becoming more common, most experiments only manipulate predator presence (Wu et al.
2011, Atwood et al. 2013, Schneider & Brose 2013), thus failing to understand the contribution of
NCEs to observed predator effects. The few studies that have investigated the role of NCEs in detrital
systems were either aquatic or focused on byproducts of predator-herbivore interactions (Steif &
Holker 2006, Boyero et al. 2008, Hawlena et al. 2012, Calizza et al. 2013).
There is clearly a need to explore NCEs in terrestrial detrital systems to understand the degree
to which control of soil properties can be attributed to the effects of predators on detritivores. This gap
in our knowledge is particularly relevant considering that predators may be more strongly linked to
detritivores than to herbivores (Wimp et al. 2013). I examined the role of CEs and NCEs in a detrital
system using the predatory wolf spider Pardosa milvina and the detritivorous collembolan Sinella
curviseta. Collembola are frequently consumed by wolf spiders, and can alter soil carbon and nutrient
dynamics (Filser 2002, Johnson et al. 2005). Specifically, collembolans can increase CO2 flux (Filser
2002, Fox et al. 2006, Kurakov et al. 2006) and soil nitrogen (Filser 2002, Kurakov et al. 2006, Pieper
& Weigmann 2008), so I predicted that interactions between predators and detritivores would cascade
to dampen these effects and that NCEs would be comparable to CEs.
62
METHODS
Additional methods and results available in appendix
I used four treatments to examine how consumptive and non-consumptive interactions affect
soil CO2 flux and nitrogen content: blank treatment (B) did not receive any arthropods and served as a
control, detritivore treatment (D) was identical to B except for the addition of 15 detritivores, cue
treatment (C) was identical to D except that it contained cues (i.e., silk, feces, and other excreta)
deposited by a single predator over a 24h period prior to the removal of the predator, and predation
treatment (P) was identical to C except that the predator was not removed before adding detritivores.
Experiments were conducted in laboratory microcosms.
I quantified daily CO2 flux for four days, and at the end of this period I removed and counted all
remaining detritivores before I analyzed soil content (nitrogen, carbon, organic carbon, C:N). I isolated
the impact of detritivores on CO2 flux and soil content by subtracting the mean values of the blank
treatment from the values measured in the detritivore treatment. I performed similar corrections for the
cue and predation treatments by subtracting the mean values from the detritivore treatment from both
the cue and predation treatments.
I tested treatment effects on the proportion of detritivores recovered at the end of the experiment
and on soil content using separate one-way ANOVAs. Flux in CO2 was analyzed using repeated
measures ANOVA. I also used a one-way ANOVA to analyze CO2 flux on the last day of the
experiment, as this represents the cumulative effect of the treatments and coincides with measurements
of soil contents. All analyses were conducted on unmanipulated and corrected values (see above), and
Welch's tests were used instead of ANOVA when groups had unequal variances. Additionally, I
calculated Cohen's d and 95% confidence intervals, using suggested guidelines to interpret effect sizes
(small = 0.2, medium = 0.5, large = 0.8) (Cohen 1988). All analyses were carried out using JMP (v9.0;
SAS Institute, Inc., Cary, NC, USA).
63
RESULTS
Detritivore survival
Predators consumed detritivores, as only 61.7% + 4.1 (mean + SE) of the detritivores survived
in the predator treatment; whereas in the absence of a predator, detritivore survival was high
(detritivore treatment: 97.4% + 0.8, cue treatment: 94.4% + 1.3). Statistically significant differences
between treatments were driven by the mortality imposed by predators, as cues alone had only a weak
effect on detritivore survival (Table 1).
CO2 flux
All treatments started at a similar state and fluctuated over time, creating an interaction between
time and treatment (F = 23.83,54, p < 0.01) with no overall treatment effect (F = 0.52,55, p = 0.62) (Figure
1). Differences between treatments were greatest on the last day of the experiment, and corrected
values revealed an increase in CO2 flux from detritivores that was absent in the predation and cue
treatments (Figure 2a, Table 2).
Soil content
I found an effect of detritivore activity on soil nitrogen, representing a 6% increase compared to
the blank treatment (Figure 2b). As predicted, adding either predator cues or an actively foraging
predator had cascading effects on the soil; nitrogen values in the predation and cue treatments were
intermediate between the blank and detritivore treatments. Corrected values illustrate the impact of
detritivores on soil nitrogen and how both predation and cues alone moderate that effect (Figure 2b,
Table 2). There was no significant effect of treatment on soil carbon or organic carbon, thus changes in
C:N were driven by effects on nitrogen.
64
DISCUSSION
I have demonstrated that predators can indirectly affect soil properties via consumptive and
nonconsumptive interactions with detritivorous prey, and that the risk of predation had effects
comparable to those from actual predation. Specifically, the presence of predators or their cues lead to a
decrease in total CO2 flux as well as reduced N inputs to the soil.
Predator cues had a large impact on soil properties despite not being renewed throughout the
experiment and causing no appreciable prey mortality. Because these cues were also present in the
predator treatment, it appears that consumptive and nonconsumptive effects are not simply additive.
Furthermore, the effects of predators seem to be largely attributable to their cues alone, as demonstrated
in grazing systems (Hawlena et al. 2012). This result is significant when considering that most studies
to date investigating the impact of predators on detrital food webs have only manipulated the presence
of predators, thus lacking the ability to highlight the importance of NCEs. These indirect interactions
may be manifest as changes in prey behavior or physiology that cascade through the soil community.
The presence of collembolans has been shown to increase CO2 flux, an effect often attributed to
collembolan stimulation of fungi and bacteria (Fox et al. 2006, Pieper & Weigmann 2008). Reduced
CO2 flux in the predator and cue treatments likely reflects decreased detritivore activity, a common
response of prey to the presence of predators and the fear of predation (Preisser et al. 2005). Indeed,
collembolans are capable of altering activity in response to predators and their cues (Nilsson &
Bengtsson 2004). Induced reductions in activity could consequently decrease stimulation of microbe
respiration, creating a system wherein predation or fear of predation cascades through prey to the soil
microbe community, ultimately altering soil processes. This system appears to have a relatively simple
structure (e.g., spiders-collembolans-microbes), as predators can reduce CO2 flux in odd-numbered
food chains and, conversely, are expected to increase flux in even-numbered food chains (Atwood et al.
2013).
Increased soil nitrogen content in response to adding detritivores is attributable to a combination
of direct inputs and interactions with soil microbes (Pieper & Weigmann 2008). Because all
collembolans were removed prior to quantifying soil content, nitrogen increases are limited to
substances left behind by individuals. Collembolans excrete N and may egest N-containing compounds
as well (Larsen et al. 2009), and, because adults continue to molt (Waldorf 1971), exuviae may also
contribute nitrogen. Sinella curviseta is particularly fecund, and deposition of spermatophores by males
and eggs by females likely contributed to increased soil nitrogen. Importantly, the nitrogen content of
65
collembolan eggs is largely derived from body reserves, not diet, providing another potential source for
the observed increase in nitrogen (Larsen et al. 2009). Finally, collembolans can increase N-fixation by
interacting with free-living N-fixers that are abundant in soil systems (Kurakov et al. 2006). The
presence of predators or their cues may have reduced these nitrogen inputs by consuming individuals or
changing detritivore behavior (Wu et al. 2011), metabolic rate (Larsen et al. 2009), assimilation
efficiency (Thaler et al. 2012), or reducing reproductive inputs. These mechanisms are not mutually
exclusive and require further study to elucidate the impacts of predators on detrital systems.
In conclusion, predators can have both consumptive and nonconsumptive effects on
detritivorous prey with cascading impacts on soil content and function. The importance of these
indirect connections is twofold because declines in biodiversity disproportionately affect predators
(Duffy 2002), and soils are important regulators of the global carbon cycle (Allison 2006).
Anthropogenic disturbances may weaken or eliminate links between predators and detritivores, with
potentially negative consequences for numerous ecosystem services provided by soil arthropods
(Lavelle et al. 2006). Studies conducted at broader spatio-temporal scales will further enhance our
understanding of the influence predators can have on detrital systems.
ACKNOWLEDGEMENTS
I am grateful to my research group, Melany Fisk, and Michael Vanni for assistance. Funding
provided by Miami University's Department of Biology and Hamilton Campus and Arachnological
Research Fund grant from the American Arachnological Society.
66
REFERENCES
Allison SD. 2006. Brown ground: a soil carbon analogue for the green world hypothesis? The
American Naturalist 167, 617-625 (doi:10.1086/503443)
Atwood TB, Hammill E, Greig HS, Kratina P, Shurin JB, Srivastava DS, Richardson JS. 2013.
Predator-induced reduction of freshwater carbon dioxide emissions. Nature Geoscience 6, 191194 (doi:10.1038/NGEO1734)
Boyero L, Rincón PA, Pearson RG. 2008. Effects of a predatory fish on a tropical detritus-based food
web. Ecologial Research 23, 649-655 (doi:10.1007/s11284-007-0424-6)
Calizza E, Rossi L, Costantini ML. 2013. Predators and resources influence phosphorus transfer along
an invertebrate food web through changes in prey behaviour. PLoS ONE 8, e65186
(doi:10.1371/journal.pone.0065186)
Cohen J. 1988. Statistical power analysis for the behavioral sciences, second edition. Hillsdale, New
Jersey: Lawrence Erlbaum Associates.
Duffy JE. 2002. Biodiversity and ecosystem function: the consumer connection. Oikos 99, 201-219
(doi:10.1034/j.1600-0706.2002.990201.x)
Filser J. 2002. The role of Collembola in carbon and nitrogen cycling in soil. Pedobiologia 46, 234-245
(doi:10.1016/S0031-4056(04)70139-0)
Fox O, Vetter S, Ekscchmitt K, Wolters V. 2006. Soil fauna modifies the recalcitrance-persistence
relationship of soil carbon pools. Soil Biology & Biochemistry 38, 1353-1363
(doi:10.1016/j.soilbio.2005.10.014)
Hassall M, Adl S, Berg M, Griffiths B, Scheu S. 2006. Soil fauna-microbe interactions: towards a
conceptual framework for research. European Journal of Soil Biology 42, S54-S60
(doi:10.1016/j.ejsobi.2006.07.007)
Hawlena D, Strickland MS, Bradford MA, Schmitz OJ. 2012. Fear of predation slows plant-litter
decomposition. Science 336, 1434-1438 (doi:10.1126/science.1220097)
Johnson D, Krsek M, Wllington EMH, Stott AW, Cole L, Bardgett RD, Read DJ, Leake JR. 2005. Soil
invertebrates disrupt carbon slow through fungal networks. Science 309, 1047
(doi:10.1126/science.1114769)
Kurakov AV, Davydova MA, Byzov BA. 2006. Microarthropods as regulators of the communities of
microscopic fungi and biological activity in the litter of a mixed forest. Eurasian Soil Science
39, 838-847 (doi:10.1134/S1064229306080060)
67
Larsen T, Ventura M, Damgaard C, Hobbie EA, Krogh PH. 2009. Nutrient allocations and metabolism
in two collembolans with contrasting reproduction and growth strategies. Functional Ecology
23, 745-755 (doi:10.1111/j.1365-2435.2009.01564.x)
Lavelle P, Decaens T, Aubert M, Barot S, Blouin M, Bureau F, Margerie P, Mora P, Rossi J-P. 2006.
Soil invertebrates and ecosystem services. European Journal of Soil Biology 42, S3-S15
(doi:10.1016/j.ejsobi.2006.10.002)
Nilsson E, Bengtsson G. 2004. Death odour changes movement pattern of a Collembola. Oikos 104,
509-517 (doi:10.1111/j.0030-1299.2004.12921.x)
Pieper S, Weigmann G. 2008. Interactions between isopods and collembolans modulate the
mobilization and transport of nutrients from urban soils. Applied Soil Ecology 39, 109-126
(doi:10.1016/j.apsoil.2007.11.012)
Preisser EL, Bolnick DI, Benard ME. 2005. Scared to death? The effects of intimidation and
consumption in predator-prey interactions. Ecology 86, 501-509 (doi:10.1890/04-0719)
Schmitz OJ. 2010. Resolving ecosystem complexity. Princeton, New Jersey: Princeton University Press.
Schneider FD, Brose U. 2013. Beyond diversity: how nested predator effects control ecosystem
functions. Journal of Animal Ecology 82, 64-71 (doi:10.1111/1365-2656.12010)
Stief P, Hölker F. 2006. Trait-mediated indirect effects of predatory fish on microbial mineralization in
aquatic sediments. Ecology 87, 3152-3159 (doi:10.1890/00129658(2006)87[3152:TIEOPF]2.0.CO;2)
Thaler JS, McArt SH, Kaplan I. 2012. Compensatory mechanisms for ameliorating the fundamental
trade-off between predator avoidance and foraging. Proceedings of the National Academies of
Sciences of the United States of America 109, 12075-12080 (doi:10.1073/pnas.1208070109)
Waldorf E. 1971. Selective egg cannibalism in Sinella curviseta (Collembola: Entomobryidae).
Ecology 52, 673-675 (doi:10.2307/1934158)
Werner EE, Peacor SD. 2003. A review of trait-mediated indirect interactions in ecological
communities. Ecology 84, 1083-1100 (doi:10.1890/00129658(2003)084[1083:AROTII]2.0.CO;2)
Wimp GM, Murphy SM, Lewis D, Douglas MR, Ambikapathi R, Van-Tull L, Gratton C, Denno RF.
2013. Predator hunting mode influences patterns of prey use from grazing and epigeic food
webs. Oecologia 171, 505-515 (doi:10.1007/s00442-012-2435-4)
Wu X, Duffy JE, Reich PB, Sun S. 2011. A brown-world cascade in the dung decomposer food web of
68
an alpine meadow: effects of predator interactions and warming. Ecological Monographs 81,
313-328 (doi:10.1890/10-0808.1)
69
Table 1. Effects of cues and predation on the survival of detritivores. Treatments: cues (C), predation
(P), detritivore (D). Symbols between treatment letters indicate relationships based on effect sizes.
Cohen's d
95% CI
Fdf
p
C=D
-0.4
(-0.9, 0.1) 37.52,65.2 <0.01
P<D
-1.8
(-2.4, -1.2)
Sample size: D (26), C (43), P (39)
70
Table 2. Effects on corrected CO2 flux and soil N content. Treatments: blank (B), cues (C), predation
(P), detritivore (D). Symbols between treatment letters indicate relationships based on effect sizes.
CO2 (mL 24h-1)
Cohen's d
95% CI
% nitrogen
Fdf
p
Cohen's d
95% CI
C<D
-0.7
(-1.3, 0.0) 13.62,27.5 <0.01
C=D
-0.5
(-1.1, 0.2)
P<D
-0.7
(-1.3, 0.0)
P=D
-0.4
(-1.0, 0.3)
C=P
-0.1
(-0.7, 0.5)
C=P
0.1
(-0.5, 0.7)
Sample size: B (20), D (21), C (20), P (17)
Fdf
p
12.52,54 <0.01
Sample size: B (20), D (20), C (19), P (18)
71
Figure 1. Corrected CO2 flux dynamics (mean + SE).
72
Figure 2. Corrected total CO2 flux on the last day of the experiment (A) and soil nitrogen (B). Box plots
show median, first and third quartiles, greatest values within 1.5 interquartile range, and outliers.
73
Chapter 4: Nonconsumptive Predator-Prey Interactions: Sensitivity of a Detritivore to
Cues of Predation Risk
74
ABSTRACT
Predators can affect prey indirectly when prey respond to cues indicating a risk of predation by altering
activity levels. Changes in prey behavior may cascade through the food web to influence ecosystem
function. I tested the response of the collembolan Sinella curviseta to cues indicating predation risk
(necromones and cues from the wolf spider Pardosa milvina). Additionally, I paired necromones and
predator cues in a conditioning experiment to determine whether the collembolan could form learned
associations. Although collembolans did not alter activity levels in response to predator cues, numerous
aspects of behavior differed in the presence of necromones. There was no detectable conditioned
response to predator cues after pairing with necromones. These results provide insight into how
collembolans perceive and respond to predation threats that vary in information content. Previously
detected indirect impacts of predator cues on ecosystem function (chapter 3) are likely due to changes
in prey other than activity level.
75
INTRODUCTION
Predation is a ubiquitous evolutionary pressure that sculpts numerous aspects of prey biology. A
realization from studies of predator-prey interactions has been that indirect predator effects (i.e., ways
predators affect prey without consuming them) have large impacts in many food webs (Werner &
Peacor 2003). Importantly, a meta-analysis comparing the direct, consumptive effects (CEs) predators
have on prey and indirect, nonconsumtive effects (NCEs) revealed that NCEs often have a larger
impact than CEs on both prey and prey resources (Preisser et al. 2005). These NCEs can negatively
impact prey foraging, activity, fecundity, and survival (Preisser & Bolnick 2008).
Placing NCEs in the context of food webs reveals an important, indirect connection between
predators and primary producers. Schmitz (2008) demonstrated that changes in grasshopper habitat use
and foraging in response to the presence of spider predators creates a system where predators have
indirect control over ecosystem function. Trophic cascades such as these are likely driven by
chemotactile cues (i.e., silk, feces, and other excreta), as these cues are sufficient for reducing
herbivory (Hlivko & Rypstra 2003, Rypstra & Buddle 2013). Although indirect effects of predators in
“green”, herbivory-based systems are well established, less attention has been given to “brown” detrital
systems. This is surprising considering the potential for parallel mechanisms to exist in both food webs
(Hassall et al. 2006), and evidence for stronger connections between predators and brown over green
food webs (Miyashita et al. 2003, Sanders et al. 2011, Wimp et al. 2013).
The few studies that have examined NCEs in detrital systems have consistently revealed
indirect impacts of predators on ecosystem function (Steif & Holker 2006, Calizza et al. 2013, Chapter
3). In the only terrestrial example, both CEs and NCEs resulted in decreased soil respiration and
nitrogen content compared to a treatment lacking predators or their cues (Chapter 3). Changes in CO2
flux and nitrogen levels are likely reflections of decreased detritivore activity, a behavioral response
that could be elicited by cues of predation risk such as silk or necromones (i.e., cues released from dead
or injured conspecifics; Yao et al. 2009). Prey commonly respond to cues of predation risk via activity
reduction (Preisser et al. 2007), which can reduce predator encounter rate and thus improve survival
(Ernsting & Jansen 1978, Persons et al. 2001). I conducted experiments to determine whether reduced
detritivore activity in response to cues of predation risk could explain previously observed reductions in
CO2 flux and soil nitrogen content. I predicted that detritivores would decrease their activity levels in
the presence of both predator cues and necromones.
76
Study species
Pardosa milvina (Araneae: Lycosidae, (Hentz)) is a common wolf spider that can reach high
densities in SW Ohio (Marshall et al. 2000). Although wolf spiders do not build webs, they deposit
information-rich chemotactile cues (hereafter referred to as cues) as they move through the
environment that can be used by their prey adaptively (Persons & Rypstra 2001, Rypstra et al. 2007,
Sitvarin & Rypstra 2012). Sinella curviseta (Collembola: Entomobryidae, Brook) is a widespread
detritivore capable of responding to sex- and density-related cues (Waldorf 1971a,b; 1974). Other
collembolans have been shown to respond to cues indicating food sources and predation risk (Negri
2004; Nilsson & Bengtsson 2004a,b; Auclerc et al. 2010; Verdeny-Vilata & Moya-Larano 2014).
Collembolans are frequently consumed by spiders (Kuusk & Eckbom 2010) and interactions between
these groups can impact ecosystem function (Lawrence & Wise 2000, 2004).
77
METHODS
Organism collection and maintenance
Pardosa milvina (study species hereafter referred to by genus) were collected from agricultural
fields at Miami University's Ecology Research Center (39°31′33′′ N, 84°43′20′′W) and maintained in
an environmental chamber with a 13:11 light:dark cycle at 25ºC. Spiders were housed individually in
plastic containers (7cm wide x 6cm tall) with a substrate of moistened peat moss and potting soil (1:1
mixture) and provided two appropriately-sized crickets (Acheta domesticus (Orthoptera: Gryllidae,
Linnaeus)) weekly. Some spiders were used as a source of cues (i.e., silk, feces, and other excreta)
more than once, but not within a period of at least two weeks between trials. Sinella were derived from
a laboratory culture and reared communally under similar conditions, but with food provided in the
form of sliced potato and baker's yeast. Only large (approximately 2mm) collembolans were used in
experiments, and individuals were only re-used in the conditioning experiment.
Response to spider cues
I evaluated the response of collembolans to spider cues by lining the bottom of a Petri dish
(4.3cm diameter) with black construction paper and allowing a single spider to deposit cues on half of
the arena for 24h. The other half of the arena was left blank, and was covered to prevent cue deposition
by the spider. After 24h, I removed the spider and uncovered the blank half of the arena. I alternated the
side of the arena with cues between trials for each experiment.
I introduced an individual collembolan into the center of the arena and allowed 30s of
acclimation before I began remotely quantifying activity for 10min using a camera mounted 1m above
the arena. I used automated motion-tracking software (EthoVision XT Version 8.0, Noldus Information
Technology, Wageningen, The Netherlands) to analyze activity levels. For each side of the arena, I
recorded the distance traveled (cm), frequency and duration (s) of time spent immobile, mobile, and
highly mobile (defined as 0%, 20%, and 60% changes in body position between frames, respectively),
turn angle (degrees), mean and total meander (turn angle/distance traveled), and velocity (distance
traveled/time between frames). When collembolans spent significantly more time on one side of the
arena, all time-sensitive variables (i.e., immobile duration, mobile duration, highly mobile duration,
and velocity) were corrected to represent proportion of time spent in those activities per side. I only
used data from trials during which collembolans visited both sides of the arena, and each individual
was used only once (n = 32).
78
I compared the time spent on each side of the arena using a paired t-test, and the activity
variables were analyzed using principal components analysis. I retained principal components with
eigenvalues greater than one and subsequently used a paired t-test on each principal component to
examine changes in collembolan behavior in response to spider cues. All analyses were completed
using R (R Core Team 2013).
Response to necromones
I tested the response of collembolans to their necromones using a split-arena design with
necromones on one half of the arena instead of spider cues. Necromones were applied by using a glass
rod to crush two collembolans and distribute their remains across half the arena. Collembolan activity
in this experiment was recorded for 8min following a 30s acclimation (n = 25). All other details are
identical to the previously described experiment.
Because I found no evidence of collembolans changing activity in response to spider cues (see
Results), I used a follow-up experiment to assess whether responses to predator cues are dependent on
a prior association between predator cues and evidence of conspecific mortality.
Conditioning
1. Pre-conditioning
To determine whether collembolans could be conditioned to respond to spider cues, I conducted
a conditioning experiment that paired a conditioned stimulus (predator cues) with an unconditioned
stimulus (necromones). I first tested the response of two groups (n = 21 per group) of collembolans to
spider cues using the same methods described in the first experiment. Spider cues were collected as
described above. Statistical tests were conducted as previously described in addition to evaluating
differences between groups using ANOVA.
2. Conditioning
I conditioned collembolans by pairing a conditioned stimulus (predator cues) with an
unconditioned stimulus (necromones). Immediately after the pre-conditioning exposure, collembolans
were placed into a new Petri dish lined with construction paper. Individuals in the conditioning group
were simultaneously exposed to spider cues (deposited as described previously) and necromones from
79
four collembolans (applied as described previously). Both cues were applied to the entire arena, and
spider cues were applied first to avoid any effects necromones might have on spider behavior.
Individuals in the control group were exposed to filter paper devoid of cues. This conditioning period
lasted for one hour, after which individuals were used again in the post-conditioning experiment.
3. Post-conditioning
I evaluated the success of the conditioning treatment by repeating the pre-conditioning
experiment with the control and conditioned groups of collembolans. I conducted the post-conditioning
experiment immediately after the conditioning period. Sample sizes were reduced to 20 (control group)
and 18 (conditioned group) due to individuals that escaped over the course of the conditioning
experiment. Statistical analyses were conducted as previously described in the pre-conditioning
experiment.
80
RESULTS
Response to spider cues
Collembolans spent equal amounts of time on the blank side of the arena and the side with
spider cues (t = -0.93, p = 0.36). Three principal components summarized their behavior (Table 1), but
there was no effect of spider cues on collembolan activity (PC1: t = -0.05, p = 0.96; PC2: t = -1.00, p =
0.32; PC3: t = -0.70, p = 0.49) (Figure 1).
Response to necromones
Collembolans responded to necromones by increasing the time spent on that half of the arena (t
= -2.41, p = 0.02) and significantly altering aspects of their behavior. Of the four principal components
used to describe their activity (Table 2), only the first two provided evidence of a response to
necromones (PC1: t = -2.75, p = 0.01; PC2: t = -2.50, p = 0.02; PC3: t = 0.24, p = 0.81; PC4: t = 0.12, p
= 0.92) (Figure 2). Principal component one reflects increased distance traveled and a high frequency
of changes in mobility state (i.e., frequently stopping and then moving), whereas principal component
two captures increases in meandering and immobility and decreased, slower movement. Therefore,
Sinella encountering necromones exhibited increased distance traveled, frequency of changes in
mobility state, meander, and immobility as well as exhibiting decreased, slower movement.
Conditioning
1. Pre-conditioning
Prior to conditioning, collembolans spent equal amounts of time on the blank and cue sides of
the arena (control group: t = -0.40, p = 0.70; conditioned group: t = 1.58, p = 0.13). The three principal
components summarizing their activity (Table 3) did not differ between groups (PC1: F1,40 = 0.51, p =
0.48; PC2: F1,40 = 2.92, p = 0.10; PC3: F1,40 = 3.84, p = 0.06), and there was no response to spider cues
for either the control group (PC1: t = -0.18, p = 0.86; PC2: t = -1.55, p = 0.14; PC3: t = 0.68, p = 0.50;
Figure 3a) or the conditioned group (PC1: t = 1.07, p = 0.30; PC2: t = 0.71, p = 0.48; PC3: t = -2.13, p
= 0.05; Figure 3b).
2. Post-conditioning
Neither group differed in time spent on one side of the arena or the other (control group: t =
-0.72, p = 0.48; conditioned group: t = -0.23, p = 0.82). Furthermore, the principal components
81
describing collembolan activity (Table 4) were not affected by conditioning (PC1: F1,36 = 0.33, p = 0.57;
PC2: F1,36 = 0.52, p = 0.48; PC3 (F1,36 = 0.79, p = 0.38). Collembolan activity did not differ between the
blank side of the arena and the side with spider cues for the control group (PC1: t = 0.34, p = 0.74;
PC2: t = -0.15, p = 0.88; PC3: t = 1.40, p = 0.18; Figue 4a) or the conditioned group (PC1: t = -0.61, p
= 0.55; PC2: t = 0.76, p = 0.46; PC3: t = 0.71, p = 0.49; Figure 4b).
82
DISCUSSION
I have demonstrated that Sinella are sensitive to environmental cues indicating risk of predation.
Specifically, individuals responded to necromones from conspecifics by altering numerous aspects of
their activity. However, I did not find evidence for NCEs propagated by cues from Pardosa: activity
levels on predator cues were no different from those on the untreated portion of the arena. Although
necromones and predator cues may co-occur in nature, thus providing the opportunity for Sinella to
learn to associate the stimuli, I did not detect a conditioned response.
The fact that Sinella responded to necromones but not to more direct predator cues may be a
product of the structure of the soil food web. Collembola face numerous predators in nature (Ernsting
& Joose 1974), and it therefore may be more advantageous to respond to necromones than to other cues
specific to each potential predator (Dicke & Grostal 2001, Nilsson & Bengtsson 2004s). Indeed, a
previous study of collembolan responses to different predation cues illustrated strong avoidance of
necromones by Protaphorura armata (Collembola: Onychiuridae (Tullberg)) but no changes in activity
when individuals encountered cues from predatory mites (Nilsson & Bengtsson 2004a). Protaphorura
increased movement speed and decreased meander in response to necromones, and these behaviors
may increase survival as the species lacks a well-developed furcula. In contrast, Sinella moved more
slowly and in a more meandering pattern when exposed to necromones. These changes in activity may
represent searching behavior, as some collembolans are known to aggregate in response to cues of
predation risk (Negri 2004). In the study by Nilsson & Bengtsson (2004a) residence time on differently
treated sections of the arena were not reported, so apparent repulsion from necromones may not have
been correctly interpreted. After correcting for differences in residence time in this study, both species
showed increased distance traveled on necromones, a response inconsistent with repulsion.
Because Sinella possess a fully-functional furcula and are capable of jumping a considerable
distance, increased activity in the presence of necromones may not be advantageous. In fact, increased
activity would likely increase the risk of predation, as visually-orienting predators are attracted to
moving prey (Ernsting & Jansen 1978, Persons & Uetz 1997). Additionally, future experiments may
reveal a priming effect (sensu Rypstra et al. 2009) of predator cues: Sinella on cues may jump sooner or
farther than those on unmanipulated substrates and thus have a survival advantage when cues are
present.
The apparent lack of learned association between necromones and predator cues in the
conditioned group may be based on the reliability of each cue type. Prey may interact with a diversity
83
of predation cues that vary in reliability (Lima & Steury 2005), especially chemical cues from predators
(Ferrari et al. 2009). The cues deposited by spiders may be ambiguous, conveying information about
the presence of the predator but not its intent. Unless prey have sophisticated sensory capabilities to
glean information about how recently a predator was present (Barnes et al. 2002) or the predator's
feeding motivation (Bell et al. 2006), predator cues may not represent information used to modify
behavior. Basing behavioral decisions on ambiguous cues could be costly for prey (Ferrari et al. 2010),
so the risk of predation may be minimized by only responding to reliable cues (e.g., necromones).
Additionally, volatility of predator cues could eliminate any potential differences in behavior on one
side of the arena versus the other (Nilsson & Bengtsson 2004a). While it is possible that learning is
simply beyond the cognitive capacity of Sinella, a large diversity of invertebrates have been
documented to display learning to some degree, even those with simple neural structures (Perry et al.
2013).
Because Sinella only responded to necromones, the impacts of predator cues on CO2 flux and
soil nitrogen content documented in Chapter 3 cannot be attributed to reductions in activity. However,
Sinella are capable of altering reproductive output in response to chemical cues (Waldorf 1971a), and a
wide diversity of species are known to reduce mating activity when under risk of predation (Lima &
Dill 1990, Krupa & Sih 1998, Hoefler et al. 2008, Fowler-Finn & Hebets 2011). Furthermore, the
observed impacts on CO2 flux and soil nitrogen may be a product of altered foraging or metabolic
processes by Sinella in response to predator cues. Specifically, prey reduce foraging when under risk of
predation (Ernsting & Jansen 1978, Preisser & Bolnick 2008) and elevate both metabolic rate and
assimilation efficiency (Hawlena & Schmitz 2010, Thaler et al. 2012). Nitrogen output from prey
sensing predator cues has been shown to increase (Hawlena & Schmitz 2010) and decrease (Thaler et
al. 2012), and it seems likely that Sinella decreases nitrogen deposition in response to predator cues
(Chapter 3).
Although predator cues do not directly alter detritivore activity, predators are still able to exert
NCEs on their prey by releasing necromones from killed conspecifics. Effects of predators on detrital
system previously attributed to CEs on detritivores (e.g., Wu et al. 2011, Schneider & Brose 2013)
likely underestimate how changes in prey behavior may contribute to processes such as decomposition.
More research into the role of NCEs in detrital systems is needed to better understand how predators
impact the brown food web, and to allow generalizations to be made encompassing detrital and
herbivory-based systems. Enhanced knowledge about detrital food webs is particularly important, as
84
soil arthropods regulate many important ecosystem services (Lavelle et al. 2006).
ACKNOWLEDGMENTS
Numerous undergraduate and graduate students helped to collect and maintain study organisms.
Sinella were derived from the Crossley Culture:
(http://www.geocities.com/fransjanssens/publicat/culture.htm). Funding was provided by Miami
University's Doctoral Undergraduate Opportunities for Scholarship award to Christian Romanchek and
me.
85
REFERENCES
Auclerc A, Libourel PA, Salmon S, Bels V, Ponge JF. 2010. Assessment of movement patterns in
Folsomia candida (Hexapoda: Collembola) in the presence of food. Soil Biology &
Biochemistry 42, 657-659 (doi:10.1016/j.soilbio.2009.12.012)
Barnes MC, Persons MH, Rypstra AL. 2002. The effect of predator chemical cue age on
antipredator behavior in the wolf spider Pardosa milvina (Araneae: Lycosidae). Journal of
Insect Behavior 15, 269-281
Bell RD, Rypstra AL, Persons MH. 2006. The effect of predator hunger on chemically mediated
antipredator responses and survival in the wolf spider Pardosa milvina (Araneae: Lycosidae).
Ethology 112, 903-910 (doi:10.1111/j.1439-0310.2006.01244.x)
Calizza E, Rossi L, Costantini ML. 2013. Predators and resources influence phosphorus transfer
along an invertebrate food web through changes in prey behaviour. PLoS ONE 8, e65186
(doi:10.1371/journal.pone.0065186)
Dicke M, Grostal P. 2001. Chemical detection of natural enemies by arthropods: an ecological
perspective. Annual Review of Ecology and Systematics 32, 1-23
Ernsting G, Joose ENG. 1974. Predation on two species of surface dwelling Collembola – study
with radioisotope labeled prey. Pedobiologia 14, 222-231
Ernsting G, Jansen JW. 1978. Interspecific and intraspecific selection by the predator Notiophilus
biguttatus F. (Carabidae) concerning two collembolan prey species. Oecologia 33, 173-183
Ferrari MCO, Sih A, Chivers DP. 2009. The paradox of risk allocation: a review and prospectus.
Animal Behaviour 78, 579-585 (doi:10.1016/j.anbehav.2009.05.034)
Ferrari MCO, Brown GE, Bortolotti GR, Chivers DP. 2010. Linking predator risk and uncertainty to
adaptive forgetting: a theoretical framework and empirical test using tadpoles. Proceeding of
the Royal Society B 277, 2205-2210 (doi:10.1098/rspb.2009.2117)
Fowler-Finn KD, Hebets EA. 2011. The degree of response to increased predation risk corresponds to
male secondary sexual traits. Behavioral Ecology 22, 268-275 (doi:10.1093/beheco/arq197)
Hassall M, Adl S, Berg M, Griffiths B, Scheu S. 2006. Soil fauna-microbe interactions: towards a
conceptual framework for research. European Journal of Soil Biology 42, S54-S60
(doi:10.1016/j.ejsobi.2006.07.007)
Hawlena D, Schmitz OJ. 2010. Herbivore physiological response to predation risk and implications for
ecosystem nutrient dynamics. Proceedings of the National Academies of Sciences of the United
86
States of America 107, 15503-15507 (doi:10.1073/pnas.1009300107)
Hlivko JT, Rypstra AL. 2003. Spiders reduce herbivory: nonlethal effects of spiders on the
consumption of soybean leaves by beetle pests. Annals of the Entomological Society of America
96, 914-919 (doi:10.1603/0013-8746(2003)096[0914:SRHNEO]2.0.CO;2)
Hoefler CD, Persons MH, Rypstra AL. 2008. Evolutionarily costly courtship displays in a wolf spider:
a test of viability indicator theory. Behavioral Ecology 19, 974-979
(doi:10.1093/beheco/arn055)
Krupa JJ, Sih A. 1998. Fishing spiders, green sunfish, and a stream-dwelling water strider: male-female
conflict and prey responses to single versus multiple predator environments. Oecologia 117,
258-265 (doi:10.1007/s004420050656)
Kuusk A-K, Ekbom B. 2010. Lycosid spiders and alternative food: feeding behavior and
implications for biological control. Biological Control 55, 20-26
(doi:10.1016/j.biocontrol.2010.06.009)
Lavelle P, Decaens T, Aubert M, Barot S, Blouin M, Bureau F, Margerie P, Mora P, Rossi J-P. 2006.
Soil invertebrates and ecosystem services. European Journal of Soil Biology 42, S3-S15
(doi:10.1016/j.ejsobi.2006.10.002)
Lawrence KL, Wise DH. 2000. Spider predation on forest-floor Collembola and evidence for indirect
effects on decomposition. Pedobiologia 44, 33-39 (doi:10.1078/S0031-4056(04)70026-8)
Lawrence KL, Wise DH. 2004. Unexpected indirect effect of spiders on the rate of litter disappearance
in a deciduous forest. Pedobiologia 48, 149-157 (doi:10.1016/j.pedobi.2003.11.001)
Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review and
prospectus. Canadian Journal of Zoology 68, 619-640 (doi:10.1139/z90-092)
Lima SL, Steury TD. 2005. Perception of predation risk: the foundation of nonlethal predator-prey
interactions. Ecology of predator-prey interactions (ed. By P. Barbosa and I. Castellanos), pp.
166-188. Oxford University Press, Oxford, U.K.
Marshall SD, Walker SE, Rypstra AL. 2000. A test for a differential colonization and competitive
ability in two generalist predators. Ecology 81, 3341-3349
Miyashita T, Takada M, Shimazaki A. 2003. Experimental evidence that aboveground predators are
sustained by underground detritivores. Oikos 103, 31-36 (doi:10.1034/j.16000706.2003.12586.x)
Negri I. 2004. Spatial distribution of Collembola in presence and absence of a predator. Pedobiologia
87
48, 585-588 (doi:10.1016/j.pedobi.2004.07.004)
Nilsson E, Bengtsson G. 2004a. Death odour changes movement pattern of a Collembola. Oikos 104,
509-517 (doi:10.1111/j.0030-1299.2004.12921.x)
Nilsson E, Bengtsson G. 2004b. Endogenous free fatty acids repel and attract Collembola. Journal of
Chemical Ecology 30, 1431-1443 (doi:10.1023/B:JOEC.0000037749.75695.c5)
Perry CJ, Barron AB, Cheng K. 2013. Invertebrate learning and cognition: relating phenomena to
neural substrate. Wiley Interdisciplinary Reviews – Cognitive Science 4, 561-582
(doi:10.1002/wcs.1248)
Persons MH, Uetz GW. 1997. The effect of prey movement on attack behavior and patch residence
decision rules of wolf spiders (Araneae: Lycosidae). Journal of Insect Behavior 10, 737-752
(doi:10.1007/BF02765390)
Persons MH, Rypstra AL. 2001. Wolf spiders show graded antipredator behavior in the presence of
chemical cues from different sized predators. Journal of Chemical Ecology 27, 2493-2504
(doi:10.1023/A:1013679532070)
Persons MH, Walker SE, Rypstra AL, Marshall SD. 2001. Wolf spider predator avoidance tactics and
survival in the presence of diet-associated predator cues (Araneae: Lycosidae). Animal
Behaviour 61, 43-51 (doi:10.1006/anbe.2000.1594)
Preisser EL, Bolnick DI, Benard ME. 2005. Scared to death? The effects of intimidation and
consumption in predator-prey interactions. Ecology 86, 501-509 (doi:10.1890/04-0719)
Preisser EL, Bolnick DI. 2008. The many faces of fear: comparing the pathways and impacts of
nonconsumptive predator effects on prey populations. PLoS ONE 3, e2465
(doi:10.1371/journal.pone.0002465)
Preisser EL, Orrock JL, Schmitz OJ. 2007. Predator hunting mode and habitat domain alter
nonconsumptive effects in predator-prey interactions. Ecology 88, 2744-2751 (doi:10.1890/070260.1)
R Development Core Team. 2013. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Rypstra AL, Schmidt JM, Reif BD, DeVito J, Persons MH. 2007. Tradeoffs involved in site selection
and foraging in a wolf spider: effects of substrate structure and predation risk. Oikos 116, 853863 (doi:10.1111/j.2007.0030-1299.15622.x)
Rypstra AL, Schlosser AM, Sutton PL, Persons MH. 2009. Multimodal signalling: the relative
88
importance of chemical and visual cues from females to the behaviour of male wolf spiders.
Animal Behaviour 77, 937-947 (doi:10.1016/j.anbehav.2008.12.026)
Rypstra AL, Buddle CM. 2013. Spider silk reduces insect herbivory. Biology Letters 9, 20120948
(doi:10.1098/rsbl.2012.0948)
Sanders D, Schaefer M, Platner C, Griffiths GJK. 2011. Intraguild interactions among generalist
predator functional groups drive impact on herbivore and decomposer prey. Oikos 120, 418-426
(doi:10.1111/j.1600-0706.2010.18924.x)
Schmitz OJ. 2008. Effects of predator hunting mode on grassland ecosystem function. Science 319,
952-954 (doi:10.1126/science.1152355)
Schneider FD, Brose U. 2013. Beyond diversity: how nested predator effects control ecosystem
functions. Journal of Animal Ecology 82, 64-71 (doi:10.1111/1365-2656.12010)
Sitvarin MI, Rypstra AL. 2012. Sex- specific response of Pardosa milvina (Araneae: Lycosidae) to
experience with a chemotactile predation cue. Ethology 118, 1230-1239 (doi:10.1111/eth.12029)
Stief P, Hölker F. 2006. Trait-mediated indirect effects of predatory fish on microbial mineralization in
aquatic sediments. Ecology 87, 3152-3159 (doi:10.1890/00129658(2006)87[3152:TIEOPF]2.0.CO;2)
Thaler JS, McArt SH, Kaplan I. 2012. Compensatory mechanisms for ameliorating the fundamental
trade-off between predator avoidance and foraging. Proceedings of the National Academies of
Sciences of the United States of America 109, 12075-12080 (doi:10.1073/pnas.1208070109)
Verdeny-Vilata O, Moya-Laraño J. 2014. Seeking water while avoiding predators: moisture gradients
can affect predator-prey interactions. Animal Behaviour 90, 101-108
(doi:10.1016/j.anbehav.2014.01.027)
Waldorf ES. 1971a. Oviposition inhibition in Sinella curviseta (Collembola: Entomobryidae).
Transactions of the American Microscopical Society 90, 314-325
Waldorf ES. 1971b. The reproductive biology of Sinella curviseta (Collembola: Entomobryidae) in
laboratory culture. Revue d'Ecologie et de Biologia du Sol 8, 451-463
Waldorf ES. 1974. Sex pheromone in the springtail, Sinella curviseta. Environmental Entomology 3,
916-918
Werner EE, Peacor SD. 2003. A review of trait-mediated indirect interactions in ecological
communities. Ecology 84, 1083-1100 (doi:10.1890/00129658(2003)084[1083:AROTII]2.0.CO;2)
89
Wimp GM, Murphy SM, Lewis D, Douglas MR, Ambikapathi R, Van-Tull L'A, Gratton C, & Denno
RF. 2013. Predator hunting mode influences patterns of prey use from grazing and epigeic food
webs. Oecologia 171, 505-515 (doi:10.1007/s00442-012-2435-4)
Wu X, Duffy JE, Reich PB, Sun S. 2011. A brown-world cascade in the dung decomposer food web of
an alpine meadow: effects of predator interactions and warming. Ecological Monographs 81,
313-328 (doi:10.1890/10-0808.1)
Yao M, Rosenfeld J, Attridge S, Sidhu S, Aksenov V, Rollo CD. 2009. The ancient chemistry of
avoiding risks of predation and disease. Evolutionary Biology 36, 267-281
(doi:10.1007/s11692-009-9069-4)
90
Table 1. Loading of activities on principal components and proportion of variation explained by each
component for the response to spider cues in the first experiment.
PC1 (55%) PC2 (21%) PC3 (13%)
Distance
0.37
0.18
-0.11
Mean meander
-0.22
0.46
-0.26
Total meander
-0.21
0.50
-0.22
Immobile time
-0.14
0.41
0.47
Immobile frequency
0.35
0.22
0.26
Mobile time
0.35
0.20
0.23
Mobile frequency
0.36
0.21
0.21
Highly mobile time
0.36
0.12
-0.24
Highly mobile frequency
0.37
0.13
-0.21
Mean turn angle
-0.18
0.37
-0.35
Velocity
0.28
-0.19
-0.52
91
Table 2. Loading of activities on principal components and proportion of variation explained by each
component for the necromone experiment.
PC1 (35%) PC2 (29%) PC3 (14%) PC4 (10%)
Distance
0.47
-0.11
0.02
-0.03
Mean meander
0.08
0.34
-0.59
0.20
Total meander
0.09
0.38
-0.48
-0.02
Immobile time
-0.04
0.50
0.34
-0.10
Immobile frequency
0.47
0.01
0.16
0.15
Mobile time
-0.11
-0.28
-0.25
0.71
Mobile frequency
0.48
-0.01
0.12
0.18
Highly mobile time
0.13
-0.43
-0.26
-0.42
Highly mobile frequency
0.44
-0.21
-0.13
-0.11
Mean turn angle
0.15
0.27
-0.26
-0.37
Velocity
-0.29
-0.31
-0.22
-0.26
92
Table 3. Loading of activities on principal components and proportion of variation explained by each
component for the response to spider cues prior to conditioning.
PC1 (44%) PC2 (19%) PC3 (18%)
Distance
0.40
-0.00
0.16
Mean meander
-0.03
-0.59
0.06
Total meander
-0.00
-0.64
0.03
Immobile time
0.11
-0.12
-0.56
Immobile frequency
0.41
0.05
-0.23
Mobile time
0.41
0.05
-0.23
Mobile frequency
0.42
0.05
-0.23
Highly mobile time
0.36
-0.11
0.29
Highly mobile frequency
0.40
-0.07
0.22
Mean turn angle
0.03
-0.45
-0.06
Velocity
0.16
0.06
0.59
93
Table 4. Loading of activities on principal components and proportion of variation explained by each
component for the response to spider cues after conditioning.
PC1 (43%) PC2 (23%) PC3 (14%)
Distance
0.39
-0.15
0.02
Mean meander
0.02
-0.59
0.03
Total meander
0.04
-0.58
-0.02
Immobile time
-0.14
-0.08
-0.60
Immobile frequency
0.39
0.06
-0.34
Mobile time
0.25
0.16
-0.28
Mobile frequency
0.37
0.06
-0.39
Highly mobile time
0.41
-0.00
0.20
Highly mobile frequency
0.44
0.03
0.02
Mean turn angle
-0.03
-0.49
-0.17
Velocity
0.33
-0.08
0.48
94
Figure 1. Collembolan activity (blank – cue) in response to spider cues in the first experiment. Box
plots show median, first and third quartiles, greatest values within 1.5 interquartile range, and outliers.
95
Figure 2. Collembolan activity (blank – cue) in response to necromones. Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
96
Figure 3. Collembolan activity (blank – cue) in response to spider cues prior to conditioning for control
(A) and experimental (B) groups. Box plots show median, first and third quartiles, greatest values
within 1.5 interquartile range, and outliers.
97
Figure 4. Collembolan activity (blank – cue) in response to spider cues after conditioning for control
(A) and experimental (B) groups. Box plots show median, first and third quartiles, greatest values
within 1.5 interquartile range, and outliers.
98
General Conclusion and Future Directions
99
Predation, a seemingly straightforward process presented to college students as a simple +/interaction between two species, is far more nuanced than initial impressions indicate. My dissertation
examines a subset of the complexity underlying predator-prey interactions by focusing on cue-mediated
behavioral shifts, prey responses to and survival with multiple predators, the importance of predatorpredator interactions, and indirect effects of predators that cascade through the food web. Although my
experiments were conducted on a handful of local species, the behavioral and ecological principles I
tested are wide in scope. Prey will likely face multiple predators in all ecosystems, and the ability to
detect and respond to predation risk is taxonomically widespread (Lima & Dill 1990, Kats & Dill 1998,
Dicke & Grostal 2001). Therefore, prey will have to integrate cues from multiple predators
representing different levels of risk and alter their behavior to reduce the risk of being captured and
consumed. Although the details of exactly how prey do this will vary from system-to-system,
fundamental concepts such as hierarchical and generalized responses (McIntosh & Peckarsky 1999)
will likely be consistently manifest.
Similarly, the framework I used to understand multiple predator effects has already been applied
to systems beyond the one in which it was developed (Schmitz 2007). Although the particular
formulation of using predator and prey characteristics (i.e., habitat domain and hunting mode) has been
criticized for being overly conservative (Tylianakis & Romo 2010), application of the underlying ideas
does help to resolve complexity in ecosystems that seems initially overwhelming. However, the
frequency and direction of intraguild predation, coupled with the existence of idiosyncratic effects on
prey suppression driven by predator identity, create a situation where fundamental natural history
knowledge of the study system is needed before predictions can be formulated, much less tested. It is
therefore important to avoid losing sight of natural history in the cloud of modern techniques (e.g.,
simulations and molecular-based approaches; Tewksbury et al. 2014). These lessons are particularly
important when considering the use of multiple predators in integrated pest management techniques, as
increased predator diversity often does not result in increased pest suppression (Griffin et al. 2013).
A further consideration when implementing predators as a means to control herbivores is that
predator connections with the detrital food web can be stronger than with the herbivory-based food web
(McNabb et al. 2001). Predator impacts in detrital food webs have received considerably less attention
than their effects on systems with primary producers at the base (Schmitz 2010), and it is only recently
that we have begun to appreciate how above-ground predators can indirectly impact below-ground
processes. The results from my small-scale, short-term experiment (chapter 3) may be indicative of
100
broader patterns in regards to predator effects on carbon dynamics (Atwood et al. 2013). The extent to
which findings from the few studies investigating these questions can be generalized remains to be
seen, though the strong conceptual parallelism with traditional, herbivory-based systems provides hope
for expansion of ecological theory to a broader range of systems (Hassall et al. 2006). Importantly, the
impacts of predators will extend beyond their consumption of prey, as alterations in prey behavior,
physiology, and interactions with other organisms are essential components driving top-down predator
effects. Although the projects described here have revealed much about complexities within and around
predator-prey interactions, there is still much work to be done. The following is a collection of ideas
that may fuel future projects, derived from what I have learned while pursuing my degree.
Chapter 1
While it is clear that we need to think about prey behavior in the context of food webs (i.e.,
consider multiple predators), there is still much work to be done. Most studies of prey behavior,
morphology, or survival in the presence of multiple predators or their cues include fewer than five
predators. This is not surprising, as adding more predators quickly increases the complexity of
experimental design and the number of replicates needed to adequately test hypotheses. That said,
increased realism is warranted, and feasibility can be maximized by identifying which predators in the
food web are likely to interact most strongly with the prey species of interest.
More specifically related to my study, greater insight into the precise mechanisms used by
Pardosa could be gained by conducting similar trials and systematically manipulating the availability
of different cue modalities. Visual cues can be eliminated by conducting trials in darkness and can
otherwise be manipulated using video playback (Clark et al. 2012), vibratory cues can be severely
dampened by using a granite substrate in place of filter paper (Gordon & Uetz 2011), and the chemical
component of chemotactile cues can be blocked by applying zinc sulfate to the legs and pedipalps of
Pardosa (Jiao et al. 2011). Based on my previous work, there may also be differences between males
and females in terms of how they respond to these cues: I would predict males to be less prone to
respond inappropriately, given the differences between sexes in memory (Sitvarin & Rypstra 2012) and
predation risk (Walker & Rypstra 2003). One possibility that has not yet been explored is that Pardosa
may exhibit similar behaviors in the presence of cues from any large arthropod at a high hunger level.
This could be a generalized response that, while maladaptive in some situations, is beneficial overall,
considering the strength of interactions between different predators in the food web. If the goal is a
101
more complete understanding of how prey sense predators and integrate potentially conflicting or
inaccurate information to make adaptive decisions, then these manipulations and experiments should
provide a productive starting framework.
Chapter 2
Although trophic cascades have been thoroughly explored and many researchers have begun
investigating the ecological consequences emergent multiple predator effects can have on prey and
their resources, there is still work to be done. The importance of intraguild predation (IGP) is poorly
understood in these systems, and no study has examined how IGP per se cascades through the food web
to affect primary producers or detritivores. Although there is a necessary tradeoff between control and
realism, multiple predator studies conducted in field enclosures will allow study species to interact with
a more complete food web than is manageable in laboratory studies. Importantly, field-based
experiments would provide the opportunity to understand how both MPEs and IGP alter the connection
strength between predators and two distinct food webs: herbivorous “green” webs and detritivorous
“brown” webs. In order to accomplish this, a prey base (i.e., herbivores or detritivores) would be
present, which may alter MPEs as mesopredators balance predator avoidance and foraging
opportunities. Of particular interest is how interactions between predators and their prey cascade to
alter decomposition, nutrient cycling, and carbon dynamics in these brown webs. Experiments such as
these will be challenging, as decompositional processes are more difficult than primary productivity
and herbivory, but they should produce interesting results that heighten our understanding of how
predators fit into their ecosystems. No long-term studies have been conducted on predator effects in
brown systems, so these should be conducted to to monitor population- and community-level
phenomena. I size-matched predators and ensured they were large enough to not be consumed by their
prey, but populations in nature are not as rigidly size-structured. Repeating the experiment with size
asymmetries between predators would likely change IGP frequency and predator identity effects.
Additionally, if the shared prey could consume its predators, the topology of the food web would
change substantially, and MPEs would have to be redefined. Finally, work from my colleagues on
herbicide impacts on spider ecology (Wrinn et al. 2012) provide a foundation for investigating how
these chemicals may alter MPEs, the frequency and direction of IGP, and the connection between
predators and the soil food web.
102
Chapter 3
My results from this study are purely phenomenological, and only scratch the surface of what is
going on in soil food webs. While arthropods may not have many direct impacts on the soil, the appear
able to indirectly regulate microbe activity. The nature of these interactions could be clarified by testing
the soil for changes in fungal enzyme profiles or even through batch sequencing, both of which would
provide information about changes in the function and composition of the soil microbe community in
response to manipulations of the arthropods present. Additionally, genetic analyses may reveal the
contribution of free-living N-fixing microbes to the soil nitrogen impacts I documented. Changes in
collembolan excretion, consumption, and reproductive output (i.e., deposition of spermatophores and
eggs) in response to predator cues could be quantified in simplified arenas using filter paper or plaster
as a substrate. Experiments such as these would provide fine-scale mechanisms that would shed light
on the cumulative effects I detected on CO2 flux and soil nitrogen content.
Thinking in the opposite direction along the scale spectrum, increasing the temporal and spatial
extent of this experiment could reveal interesting effects that can't be detected in laboratory
microcosms. Of particular interest to me are the impacts of adding a higher trophic level to the system:
simple trophic cascade theory would predict that higher-order predators would reverse the impacts on
CO2 flux and soil nitrogen seen in their absence (Atwood et al. 2013). Additionally, increasing the scale
of this experiment would allow investigation of whether top-down effects of predators translate to
bottom-up effects on plants. Finally, a more applied approach could be taken: comparing soils from
organic and traditionally-managed farms and explicitly examining how pesticide or herbicide residues
affect predator impacts on soil properties.
Chapter 4
The lack of response to cues from Pardosa seems puzzling, considering that Pardosa readily eat
Sinella as adults and immediately upon dispersing from their mothers. I did detect behavioral responses
to necromones, so it may be that Sinella would respond to spider cues if the spiders were previously
maintained on a diet of Sinella (Schoeppner & Relyea 2009, Hoefler et al. 2012). Other possible
responses to spider cues may exist that would not have been captured given the experimental design: 1)
Sinella may avoid predator cues in a three-dimensional way (Grear & Schmitz 2005) or by trying to
emigrate from the arena, 2) spider cues may prime Sinella to respond to the threat of predation by
decreasing the latency until jumping, increasing the likelihood of jumping, or increasing jump distance,
103
3) responses may be social, as collembolans may aggregate as a selfish effort to minimize risk of
predation, and finally 4) cues beyond the chemotactile information present in this experiment may elicit
antipredator behavior (Munoz & Blumstein 2012, Ben-Ari & Inbar 2014). These could include visual,
substrate-borne, or air pressure (Casas et al. 2008) cues originating from an attacking spider. Sinella are
capable of avoiding capture by Pardosa, but it is not clear how they do this. Because it is well known
that Pardosa is sensitive to chemotactile cues, it would be worthwhile to investigate changes in
Pardosa behavior in response to cues deposited by Sinella.
104
REFERENCES
Atwood TB, Hammill E, Greig HS, Kratina P, Shurin JB, Srivastava DS, Richardson JS. 2013 Predatorinduced reduction of freshwater carbon dioxide emissions. Nature Geoscience 6, 191-194
(doi:10.1038/NGEO1734)
Ben-Ari M, Inbar M. 2014. Aphids link different sensory modalities to accurately interpret ambiguous
cues. Behavioral Ecology 25, 627-632 (doi:10.1093/beheco/aru033)
Casas J, Steinmann T, Dangles O. 2008. The aerodynamic signature of running spiders. PLoS ONE 3,
e2116. (doi:10.1371/journal.pone.0002116)
Clark DL, Roberts JA, Uetz GW. 2012. Eavesdropping and signal matching in visual courtship displays
of spiders. Biology Letters 8, 375-378. (doi:10.1098/rsbl.2011.1096)
Dicke M, Grostal P. 2001. Chemical detection of natural enemies by arthropods: An ecological
perspective. Annual Review of Ecology and Systematics 32, 1-23
Grear JS, Schmitz OJ. 2005. Effects of grouping behavior and predators on the spatial distribution of a
forest floor arthropod. Ecology 86, 960-971 (doi:10.1890/04-1509)
Griffin JN, JEK Byrnes, Cardinale BJ. 2013. Effects of predator richness on prey suppression: a metaanalysis. Ecology 94, 2180-2187 (doi:10.1890/13-0179.1)
Hassall M, Adl S, Berg M, Griffiths B, Scheu S. 2006 Soil fauna-microbe interactions: towards a
conceptual framework for research. European Journal of Soil Biology 42, S54-S60
(doi:10.1016/j.ejsobi.2006.07.007)
Hoefler CD, Durso LC, McIntyre KD. 2012. Chemical-mediated predator avoidance in the European
house cricket (Acheta domesticus) is modulated by predator diet. Ethology 118, 431-437
(doi:10.1111/j.1439-0310.2012.02028.x)
Jiao X, Chen Z, Du H, Chen J, Liu F. 2011. Chemoreceptors distribution and relative importance of
male forelegs and palps in intersexual chemical communication of the wolf spider Pardosa
astrigera. Chemoecology 21, 45-49 (doi:10.1007/s00049-010-0062-2)
Kats LB, Dill LM. 1998. The scent of death: chemosensory assessment of predation risk by prey
animals. Ecoscience 5, 361-394
Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review and
prospectus. Canadian Journal of Zoology 68, 619-640 (doi:10.1139/z90-092)
McIntosh AR, Peckarsky BL. 1999. Criteria determining behavioural responses to multiple predators
by a stream mayfly. Oikos 85, 554-564 (doi:10.2307/3546705)
105
McNabb DM, Halaj J, Wise DH. 2001. Inferring trophic positions of generalist predators and their
linkage to the detrital food web in agroecosystems: a stable isotope analysis. Pedobiologia 45,
289-297 (doi:10.1078/0031-4056-00087)
Munoz NE, Blumstein DT. 2012. Multisensory perception in uncertain environments. Behavioral
Ecology 23, 457-462 (doi:10.1093/beheco/arr220)
Schmitz OJ. 2007. Predator diversity and trophic interactions. Ecology 88, 2415-2426 (doi:10.1890/060937.1)
Schmitz OJ. 2010. Resolving ecosystem complexity. Princeton, New Jersey: Princeton University Press.
Schoeppner NM, Relyea RA. 2009. Interpreting the smells of predation: how alarm cues and
kairomones induce different prey defences. Functional Ecology 23, 1114-1121
(doi:10.1111/j.1365-2435.2009.01578.x)
Sitvarin MI, Rypstra AL. 2012. Sex- specific response of Pardosa milvina (Araneae: Lycosidae)
to experience with a chemotactile predation cue. Ethology 118, 1230-1239
(doi:10.1111/eth.12029)
Tewksbury JJ, Anderson JGT, Bakker JD, Billo TJ, Dunwiddie PW, Groom MJ, Hampton SE, Herman
SG, Levey DJ, Machnicki NJ, Del Rio CM, Power ME, Rowell K, Salomon AK,Stacey L,
Trombulak SC, Wheeler TA. 2014. Natural history's place in science and society. BioScience
64, 300-310 (doi:10.1093/biosci/biu032)
Tylianakis JM, CM Romo. 2010. Natural enemy diversity and biological control: making sense of the
context-dependency. Basic and Applied Ecology 11, 657-668 (doi:10.1016/j.baae.2010.08.005)
Walker SE, Rypstra AL. 2003. Sexual dimorphism and the differential mortality model: is behaviour
related to survival? Biological Journal of the Linnean Society 78, 97-103 (doi:10.1046/j.10958312.2003.00134.x)
106
Appendix
107
CHAPTER 1: SUPPLEMENTARY RESULTS
Table A1. Mean (SE) for each activity variable used in the principal component analysis of Pardosa
response to patchy cues from Tigrosa and Scarites at two hunger levels.
Low hunger
Blank Tigrosa Scarites
High hunger
Both
Blank
Tigrosa
Scarites
Both
Distance (cm)
1136.7 552.97 1290.0 457.0
(837.7) (569.2) (985.5) (448.5)
1136.7
(837.7)
69.2
(79.6)
149.9
94.8
(105.3) (122.2)
Ambulatory (s)
428.5 270.0
457.8 317.3
(191.8) (206.7) (247.5) (245.3)
428.5
(191.8)
356.3
(291.5)
515.4
400.5
(218.1) (297.3)
Stereotypic (s)
280.2 160.2
282.9 212.0
(115.8) (113.4) (135.4) (131.5)
280.2
(115.8)
237.0
(153.6)
258.4
227.4
(113.4) (107.9)
Immobile (s)
1091.4 1368.9 1059.4 1270.8
(291.5) (306.7) (360.6) (365.9)
1091.4
(291.5)
1206.7
(428.1)
1026.3 1172.1
(317.6) (381.3)
Speed (cm/s)
1.45
(0.61)
1.45
(0.61)
0.09
(0.06)
0.17
(0.08)
1.03
(0.60)
1.51
(0.79)
0.72
(0.31)
108
0.12
(0.11)
CHAPTER 2: SUPPLEMENTARY METHODS
During June & July 2010, I conducted a laboratory study to quantify the habitat domain of
Pardosa, Tigrosa, Rabidosa, and Scarites. Approximately 1cm of moistened soil was added to each of
11 glass terraria (77cm long x 31.5cm wide x 32cm tall). The depth of the soil provided traction for
organisms and allowed Scarites to burrow without disappearing from sight. To provide an opportunity
for organisms to climb, I placed three evenly spaced bundles of straw (16cm tall) along the shortest
wall of each terrarium. Window screening was used to prevent organisms from entering or leaving
terraria. Eleven terraria were arranged on metal shelving units with opaque barriers between terraria to
prevent organisms from interacting visually. Trials were conducted in an environmental chamber
maintained at 25ºC and 60% RH with a light:dark schedule of 16h:8h, and all animals were acclimated
to these conditions for at least one week prior to being used.
All organisms were provided two crickets four days before being used. Because Tigrosa,
Rabidosa, and Scarites are most active at night, I began trials approximately 4h after the onset of the
dark cycle. Organisms were placed singly in each terrarium and allowed to acclimate for five minutes
under a transparent vial before being released onto the soil in the center of each terrarium. Following
release, I recorded the activity and habitat use of each individual once every 15 minutes. Each trial
lasted 3h, thus individuals were observed a total of twelve times over the course of each trial. To
minimize disturbance, all observations were made using a dim red light for illumination. Species were
randomly assigned to terraria, and all 11 terraria were run simultaneously once or twice a week, with at
least two replicates of each species per trial. I used 20 individuals of each species, and each individual
was used only once.
For each observation, I recorded the location of each organism within the terrarium. Changes in
vertical and horizontal position between observations were used to infer movement during the interval
between observations. To estimate activity level, the vertical position (height) and horizontal
displacement (movement between observations) of organisms at each observation were independently
summed. I defined mutually exclusive habitats as ground (on the soil surface), subterranean (below the
soil surface), vegetation (climbing on the straw), and wall (attempting to climb terrarium wall). The
number of times an individual was observed in a particular habitat was summed to estimate habitat use.
Differences in domain (habitat use and activity level) between species were examined using a KruskalWallis test on each significant component of a principal components analysis (PCA).
109
CHAPTER 2: SUPPLEMENTARY RESULTS
All species actively explored the terraria and were observed in the different habitats provided,
with the exception that only Scarites was ever observed beneath the soil. Because organisms were
rarely observed attempting to climb terrarium walls (less than 2% of observations), vertical habitat use
and vertical position were calculated regardless of habitat (i.e., no distinction made between height on
wall or straw). The principal component analysis returned two variables (PC1 and PC2) with
eigenvalues greater than one, and accounted for 56 and 22% of the variability, respectively.
Overall, species did not differ in their use of the habitats provided (Figure A1). Rabidosa was
frequently observed climbing, and thus had the highest score for principal component one, which
related positively to vertical habitat use and vertical position (Figure A1a). Tigrosa and Scarites were
rarely observed climbing, and Pardosa used the vertical habitat with intermediate frequency (Figure
A1a). However, the species did not differ overall with respect to PC1 (Kruskal-Wallis test, X2 = 3.7670,
df = 3, p = 0.2878).
Horizontal displacement and subterranean habitat use loaded positively on PC2, and the species
differed overall in this regard (Kruskal-Wallis test, X2 = 35.5612, df = 3, p < 0.0001; Figure A1b).
Scarites moved quickly throughout the terrarium, and thus had the highest horizontal displacement in
addition to the most frequent use of the subterranean habitat. Conversely, Pardosa and Tigrosa were
fairly inactive, and Rabidosa often wandered around the terrarium before beginning to climb on the
straw.
Differences between species in habitat use, activity, and prey capture are summarized in Table
A2. Pardosa is classified as having a broad domain because it is typically more active during the day
(Marshall et al. 2002, Schonewolf et al. 2006) and has been shown to move vertically to avoid
predation (Lowrie 1973, Folz et al. 2006); whereas this experiment was conducted in the dark and
without the presence of predation risk.
110
Table A2. Summary of habitat domain and hunting mode for Pardosa, Tigrosa, Rabidosa, and Scarites.
Habitat domain and hunting mode classified according to supplementary results from chapter 2,
personal observations, and published literature.
Habitat domain
Hunting mode
Pardosa
Broad
Sit-and-move
Tigrosa
Narrow (on soil)
Sit-and-move
Rabidosa
Narrow (vegetation)
Sit-and-move
Scarites Narrow (on and below soil)
Active
111
Figure A1. Characterization of habitat domain for Pardosa, Tigrosa, Rabidosa, and Scarites (mean +
SE).Vertical displacement and vertical habitat use loaded positively on PC1, whereas use of the soil
surface loaded negatively (a). Horizontal displacement and use of the subterranean habitat loaded
positively on PC2 (b).
112
CHAPTER 3: SUPPLEMENTARY METHODS
Organism collection and maintenance
I collected immature and adult female Pardosa milvina from Miami University's Ecology
Research Center (39°31′33′′ N, 84°43′20′′W). Spiders were housed individually in plastic containers
(7cm wide x 6cm tall) with a substrate of commercially available peat moss and potting soil (1:1
mixture), provided water ad libitum, and given two crickets (Acheta domesticus) once a week. All
spiders were used in experiments one week after their most recent feeding. Sinella curviseta were
derived from the Crossley Culture (http://www.geocities.com/fransjanssens/publicat/culture.htm) and
reared communally under similar conditions with the exception that food was provided in the form of a
sliced potato and baker's yeast. Only large (approx. 2mm) S. curviseta were used in experiments. All
study organisms were maintained in an environmental chamber with a 13h:11h light:dark cycle, at
25ºC. No organisms were used in more than one trial.
Experimental setup
The experiment was conducted within horizontally-oriented Mason jars (8.5cm wide x 15.5cm
tall, 947mL), each containing 60g of 1:1 mixture of peat moss:potting soil. I added 1g of dried straw
(3cm long pieces) on top of the soil to provide habitat structure and a substrate for microbe growth. Jars
were cleaned with ethanol between trials.
CO2 measurements
I placed a glass vial with 10mL 0.25M NaOH centrally in each jar immediately after adding
detritivores, and the vial was replaced with a new vial every 24h for four days. Total daily CO2 flux
was quantified by titration with 0.25M HCl to determine the quantity of CO2 absorbed following
established procedures (Fisk et al. 1998).
Soil content measurements
I dried soil and straw at 60ºC for 48h and followed established protocols (Vanni et al. 2011).
Samples were ground and analyzed for C and N content using a Flash 2000 Combustion NC soil
analyzer (CE Elantech, Inc., Lakewood, NJ, USA). Sub-samples were taken to determine organic
carbon content by ashing samples at 550ºC for 4h and subtracting post-ashed C content from pre-ashed
C content and correcting for mass lost on ignition.
113
Statistical analyses
I used simple linear regression to test the relationship between the consumption of detritivores
by predators and the total daily CO2 flux on the last day of the experiment (n = 21-32). I tested for
correlation between consumption of detritivores by predators and soil content (i.e., %C, % organic C,
%N, and C:N) using MANOVA. CO2 flux dynamics were analyzed using repeated measures ANOVA,
with treatment as a factor (n = 17-21). Differences between treatments in total C, organic C, and C:N
were analyzed with one-way ANOVA (n = 18-20). All analyses were conducted on unmanipulated and
corrected values (see main text), and Welch's tests were used instead of ANOVA when groups had
unequal variances. All analyses were carried out using JMP (version 9.0; SAS Institute, Inc., Cary, NC,
USA).
114
CHAPTER 3: SUPPLEMENTARY RESULTS
Detritivore survival
Consumption of detritivores was negatively correlated with CO2 flux on the last day of the
experiment (r = 0.8, 95% CI = 0.6-0.9; = R2 = 0.65, p < 0.01; Figure A2). The proportion of detritivores
surviving at the end of the experiment did not correlate with any measures of soil content (MANOVA:
F4,12 = 0.18, p = 0.72).
CO2 flux dynamics
All treatments started at a similar state and fluctuated over time (Figure A3), with a significant
time effect and an interaction between time and treatment despite there being no significant overall
treatment effect (Table A3). Unmanipulated CO2 flux values for the last day of the experiment illustrate
the impact of detritivores, as they significantly increased CO2 values (Figure A4, Table A4).
Soil C, organic C, and C:N content
All treatments resulted in a slight increase in soil carbon content compared to the control
(Figure A5a, Table A6). This effect was driven by the addition of detritivores, as corrected values
revealed no differences between the detritivore treatment and either the cue or predation treatment
(Figure A5b, Table A6). The carbon in the soil was over 99% organic, so the treatment effects on
organic carbon content are qualitatively the same as for total carbon (Figure A6, Table A7). Differences
between treatments in soil C:N are a product of the impacts of detritivores, predator cues, and predation
on carbon and nitrogen content, and their impacts on soil C:N were fairly uniform across treatments
(Figure A7a). The impact of detritivores on soil nitrogen drove statistical differences between
treatments for corrected C:N values (Figure A7b, Table A8).
115
CHAPTER 3: SUPPLEMENTARY REFERENCES
Fisk MC, Schmidt SK, Seastedt TR. 1998 Topographic patterns of above- and belowground production
and nitrogen cycling in alpine tundra. Ecology 79, 2253-2266 (doi:10.1890/00129658(1998)079[2253:TPOAAB]2.0.CO;2)
Vanni MJ, Renwick WH, Bowling AM, Horgan MJ, Christian AD. 2011 Nutrient stoichiometry of
linked catchment-lake systems along a gradient of land use. Freshwater Biology 56, 791-811
(doi:10.1111/j.1365-2427.2010.02436.x)
116
Table A3. Treatment effects on unmanipulated CO2 flux dynamics (repeated measures ANOVA).
Time
Treatment
Time*Treatment
Fdf
38.43,72
0.13,74
5.53,74
p
<0.01
0.95
<0.01
Sample size: B (20), D (21), C (20), P (17)
117
Table A4. Treatment effects on unmanipulated CO2 flux from the last day of the experiment. Statistics
reported are Cohen's d, 95% confidence intervals, and F-values, degrees of freedom, and p-values.
Treatments are: blank (B), cues (C), predation (P), detritivore (D). Symbols between treatment letters
indicate relationships based on effect sizes.
CO2 (mL 24h-1)
Cohen's d 95% CI
Fdf
p
D>B
0.7
(0.0, 1.3) 3.23,34.8 0.03
C=B
0.0
(-0.6, 0.6)
P=B
-0.1
(-0.7, 0.6)
Sample size: B (20), D (21), C (20), P (17)
118
Table A5. Treatment effects on unmanipulated soil N content. Statistics reported are Cohen's d, 95%
confidence intervals, and F-values, degrees of freedom, and p-values. Treatments are: blank (B), cues
(C), predation (P), detritivore (D). Symbols between treatment letters indicate relationships based on
effect sizes.
% nitrogen
Cohen's d 95% CI
Fdf
p
D>B
0.8
(0.2, 1.4) 2.83,73 0.04
C=B
0.5
(-0.2, 1.1)
P=B
0.5
(-0.1, 1.2)
Sample size: B (20), D (20), C (19), P (18)
119
Table A6. Treatment effects on soil C content. All tests were performed on unmanipulated and
corrected values (see methods). Statistics reported are Cohen's d, 95% confidence intervals, and Fvalues, degrees of freedom, and p-values. Treatments are: blank (B), cues (C), predation (P), detritivore
(D). Symbols between treatment letters indicate relationships based on effect sizes.
% carbon
Cohen's d
95% CI
% carbon (corrected)
Fdf
p
Cohen's d
95% CI
Fdf
p
D>B
0.6
(0.0, 1.3) 2.33,73 0.08
C=D
0.1
(-0.5, 0.8) 2.62,54 0.09
C>B
0.8
(0.1, 1.4)
P=D
0.0
(-0.6, 0.6)
P=B
0.6
(-0.1, 1.2)
C=P
-0.1
(-0.8, 0.5)
Sample size: B (20), D (20), C (19), P (18)
120
Table A7. Treatment effects on soil organic C content. All tests were performed on unmanipulated and
corrected values (see methods). Statistics reported are Cohen's d, 95% confidence intervals, and Fvalues, degrees of freedom, and p-values. Treatments are: blank (B), cues (C), predation (P), detritivore
(D). Symbols between treatment letters indicate relationships based on effect sizes.
% organic carbon
Cohen's d
95% CI
% organic carbon (corrected)
Fdf
p
Cohen's d
95% CI
Fdf
p
D=B
0.6
(-0.1, 1.2) 1.93,72 0.14
C=D
0.1
(-0.5, 0.8) 2.02,54 0.15
C>B
0.7
(0.0, 1.3)
P=D
0.0
(-0.6, 0.6)
P=B
0.5
(-0.1, 1.2)
C=P
-0.1
(-0.8, 0.5)
Sample size: B (19), D (20), C (19), P (18)
121
Table A8. Treatment effects on soil C:N. Statistics reported are Cohen's d, 95% confidence intervals,
and F-values, degrees of freedom, and p-values. Treatments are: blank (B), cues (C), predation (P),
detritivore (D). Symbols between treatment letters indicate relationships based on effect sizes.
C:N
Cohen's d
C:N (corrected)
95% CI
Fdf
p
Cohen's d
95% CI
Fdf
p
D=B
-0.3
(-0.9, 0.3) 1.63,73 0.20
C=D
0.6
(0.0, 1.3) 4.52,34.4 0.02
C=B
0.4
(-0.3, 1.0)
P=D
0.4
(-0.3, 1.0)
P=B
0.1
(-0.6, 0.7)
C=P
-0.4
(-1.1, 0.2)
Sample size: B (20), D (20), C (19), P (18)
122
Figure A2. Relationship between detritivores consumed by predators and total CO2 flux on the last day
of the experiment.
123
Figure A3. Unmanipulated CO2 flux dynamics (mean +SE).
124
Figure A4. Unmanipulated CO2 flux on the last day of the experiment. Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
125
Figure A5. Soil C content for both unmanipulated (A) and corrected values (B). Box plots show
median, first and third quartiles, greatest values within 1.5 interquartile range, and outliers.
126
Figure A6. Soil organic C content for both unmanipulated (A) and corrected values (B). Box plots show
median, first and third quartiles, greatest values within 1.5 interquartile range, and outliers.
127
Figure A7. Soil C:N for both unmanipulated (A) and corrected values (B). Box plots show median, first
and third quartiles, greatest values within 1.5 interquartile range, and outliers.
128
CHAPTER 4: SUPPLEMENTARY METHODS
In addition to testing the activity response of Sinella in the presence of chemotactile cues from
Pardosa (see Chapter 4), I also ran the first part of the experiment using Tigrosa, Scarites, and
Rabidosa as sources of cues. These trials were conducted simultaneously with those described in
Chapter 4, and the methods were identical except for the species used as a cue source. I analyzed the
data including trials with cues from Pardosa, though this has no impact on the outcome of the analysis.
129
CHAPTER 4: SUPPLEMENTARY RESULTS
The source of predator cues (i.e., Pardosa, Tigrosa, Rabidosa, or Scarites) did not significantly affect
any measured aspect of Sinella activity (PC1: F3,124 = 0.23, p = 0.88; PC2: F3,124 = 0.18, p = 0.91; PC3:
F3,124 = 0.08, p = 0.97; Table A9; Figures A8-10). With the exception of Pardosa, these predators are too
large to consume Sinella, so the lack of response to their cues is likely adaptive.
Mean (SE) values for activity metrics used to compute principal components appear in Tables
A10 and A11.
130
Table A9. Loadings of activity metrics on principal components and proportion of total variation
explained by each component.
PC1 (54.5%) PC2 (20.4%) PC3 (14.1%)
Distance (cm)
0.95
0.09
0.16
Immobile time (s)
-0.19
0.82
-0.23
Immobile frequency
0.84
0.42
-0.23
Mobile time (s)
0.87
0.37
-0.18
Mobile frequency
0.90
0.36
-0.16
Highly mobile time (s)
0.88
-0.15
0.35
Highly mobile frequency
0.92
-0.08
0.28
Speed (cm/s)
0.63
-0.65
0.29
Turn angle (degrees)
-0.33
0.48
0.68
Meander (degrees/cm)
-0.33
0.49
0.69
131
Table A10. Mean (SE) for each activity variable used in the principal component analysis of Sinella
response to cues from Pardosa as well as the response to necromones. Each cue source (Pardosa cues
or necromones) is paired with a blank.
Pardosa
Blank
cues
Necromones Blank
Distance (cm)
62.9
(6.3)
60.5
(6.9)
48.4
(5.3)
63.7
(6.8)
Immobile (s)
200.1
(21.4)
166.6
(20.1)
89.9
(15.5)
151.6
(17.7)
Immobile freq.
461.4
(41.7)
427.8
(36.9)
319.1
(31.4)
433.6
(38.3)
Mobile (s)
82.1
(8.0)
74.2
(6.8)
56.8
(5.9)
78.9
(6.3)
Mobile freq.
487.0
(44.3)
447.5
(40.9)
321.1
(32.2)
443.9
(39.3)
Highly mobile (s)
38.3
(5.7)
38.4
(6.2)
43.6
(5.6)
58.1
(6.7)
Highly mobile freq.
260.3
(36.1)
264.9
(41.6)
249.4
(30.5)
327.5
(40.0)
Speed (cm/s)
0.22
(0.02)
0.23
(0.02)
0.29
(0.02)
0.25
(0.02)
Turn angle (deg.)
1.01
(0.31)
0.75
(0.36)
-1.08
(0.46)
0.51
(0.42)
1358.4 813.3
(469.4) (464.4)
-56.75
(240.5)
566.5
(315.4)
Meander (deg./cm)
132
Table A11. Mean (SE) for each activity variable used in the principal component analysis of Sinella
response to cues from Pardosa before and after conditioning. Each cue source is paired with a blank.
PreBlank
conditioning
Postconditioning
Blank
Distance (cm)
53.2
(4.6)
54.36
(4.8)
69.0
(4.5)
71.8
(7.0)
Immobile (s)
120.2
(12.5)
137.9
(14.3)
130.5
(11.2)
113.5
(10.6)
Immobile freq.
393.1
(28.9)
411.8
(33.6)
300.9
(15.8)
286.4
(18.9)
Mobile (s)
69.5
(5.1)
74.1
(6.1)
52.0
(3.1)
50.5
(4.0)
Mobile freq.
406.0
(29.4)
430.1
(35.1)
316.2
(16.1)
292.7
(18.4)
Highly mobile (s)
36.1
(3.5)
38.3
(4.6)
62.3
(5.1)
64.7
(7.0)
Highly mobile freq.
227.3
(18.6)
243.7
(28.7)
285.5
(19.7)
282.6
(23.7)
Speed (cm/s)
0.26
(0.02)
0.23
(0.02)
0.32
(0.02)
0.35
(0.03)
Turn angle (deg.)
-0.35
(0.43)
0.05
(0.42)
0.53
(0.41)
-0.20
(0.36)
Meander (deg./cm)
-34.7
(214.0)
527.8
(538.0)
264.9
(213.3)
108.5
(244.9)
133
Figure A8. Collembolan activity in response to cues from Pardosa (grey), Tigrosa (brown), Rabidosa
(yellow), and Scarites (black). Box plots show median, first and third quartiles, greatest values within
1.5 interquartile range, and outliers.
134
Figure A9. Collembolan activity in response to cues from Pardosa (grey), Tigrosa (brown), Rabidosa
(yellow), and Scarites (black). Box plots show median, first and third quartiles, greatest values within
1.5 interquartile range, and outliers.
135
Figure A10. Collembolan activity in response to cues from Pardosa (grey), Tigrosa (brown), Rabidosa
(yellow), and Scarites (black). Box plots show median, first and third quartiles, greatest values within
1.5 interquartile range, and outliers.
136