CONSEQUENCES OF MUTUALISMS BETWEEN APHIDS AND 
AN INVASIVE ANT TO ARTHROPOD COMMUNITIES 
AND THEIR HOST PLANTS 
 
Except where reference is made to the work of others, the work described in this 
dissertation is my own or was done in collaboration with my advisory 
committee.  This dissertation does not include proprietary 
or classified information. 
 
 
       
John D. Styrsky 
 
 
Certificate of Approval: 
 
 
 
             
Kathy L. Flanders     Micky D. Eubanks, Chair 
Associate Professor     Associate Professor 
Entomology and Plant Pathology   Entomology and Plant Pathology 
 
 
 
             
Henry Y. Fadamiro     Sharon M. Hermann 
Assistant Professor     Assistant Visiting Professor 
Entomology and Plant Pathology   Biological Sciences 
 
 
     
          
Stephen L. McFarland 
Acting Dean 
Graduate School 
 
CONSEQUENCES OF MUTUALISMS BETWEEN APHIDS AND 
AN INVASIVE ANT TO ARTHROPOD COMMUNITIES 
AND THEIR HOST PLANTS 
 
 
John D. Styrsky 
 
A Dissertation 
Submitted to 
the Graduate Faculty of 
Auburn University 
in Partial Fulfillment of the 
Requirements for the  
Degree of 
Doctor of Philosophy 
 
 
Auburn, Alabama 
May 11, 2006 
 
 
 
iii 
CONSEQUENCES OF MUTUALISMS BETWEEN APHIDS AND 
AN INVASIVE ANT TO ARTHROPOD COMMUNITIES 
AND THEIR HOST PLANTS 
 
 
John D. Styrsky 
 
Permission is granted to Auburn University to make copies of this dissertation                
at its discretion, upon request of individuals or institutions at their expense.                           
The author reserves all publication rights. 
 
 
             
      Signature of Author 
 
             
      Date of Graduation 
 
iv 
DISSERTATION ABSTRACT 
CONSEQUENCES OF MUTUALISMS BETWEEN APHIDS AND 
AN INVASIVE ANT TO ARTHROPOD COMMUNITIES 
AND THEIR HOST PLANTS 
 
John D. Styrsky 
Doctor of Philosophy, May 11, 2006 
(M.S., Illinois State University, 1999) 
(B.S., Southwestern University, 1992) 
 
166 Typed pages 
Directed by Micky D. Eubanks 
 
Mutualisms between ants and honeydew-producing Hemipteran insects (e.g., 
aphids) are abundant and widespread in arthropod food webs, yet their ecological 
consequences are very poorly known.  Previous work in several agricultural systems in 
Alabama show that the community-level effects of red imported fire ants (Solenopsis 
invicta) vary among crops and that this variation may be correlated with the presence and 
absence of aphids.  Fire ants have stronger and more pervasive effects in cotton, for 
example, where they are attracted onto plants by cotton aphids (Aphis gossypii) than they 
do in soybean, which has not historically hosted aphids.  Alternatively, fire ants may be 
hindered from foraging on soybean plants by dense trichomes on soybean stems and 
 
v 
leaves.  In chapter one, I present field and greenhouse experiments in which I 
manipulated fire ant density in plots of three soybean isolines varying in trichome density 
to test this hypothesis.  Trichomes did not inhibit fire ants from foraging on plants in the 
field or in the greenhouse, and fire ant predation of herbivores in the field was actually 
greater on pubescent plants relative to glabrous plants.  In chapter two, I review several 
studies that investigate explicitly the consequences of ant-Hemipteran mutualisms to 
plants.  In most cases, plants benefit from these mutualisms as a consequence of 
increased predation or harassment of more damaging herbivores by Hemipteran-tending 
ants, resulting in decreased plant damage and increased plant reproduction.  In chapter 
three, I describe field and greenhouse experiments in which I manipulated the presence 
and absence of cotton aphids on cotton plants to test the hypothesis that a mutualism 
between cotton aphids and red imported fire ants benefits cotton plants by increasing fire 
ant suppression of beet armyworm caterpillars (Spodoptera exigua).  Greater numbers of 
fire ants foraged on plants with cotton aphids than on plants without cotton aphids, 
resulting in a significant reduction in caterpillar survival and caterpillar herbivory of 
leaves, squares, and bolls on plants with aphids.  In chapter four, I describe field and 
greenhouse experiments to test the hypothesis that the range expansion of the soybean 
aphid (Aphis glycines), will cause a dramatic increase in the effects of fire ants as 
predators in soybean.  In field experiments in which I misted soybean plants with 
artificial honeydew as a surrogate for soybean aphid honeydew, fire ants were 
significantly more abundant on honeydew-misted plants, resulting in decreased beet 
armyworm caterpillar abundance and caterpillar damage to plants.  Results from 
greenhouse experiments with soybean aphids were consistent with the field experiments. 
 
vi 
ACKNOWLEDGMENTS 
 
 Both my research and my development as a scientist greatly benefited from the 
guidance, motivation, and insight provided by my adviser and friend, Micky Eubanks.  
His enthusiasm and generosity with his time contributed in no small way to the success of 
this work.  I sincerely thank the other members of my committee, Henry Fadamiro, Kathy 
Flanders, and Sharon Hermann, and my outside reader, Robert Boyd, for their time and 
genuine interest.  Tom Barnum, Michael Buckman, Heidi Connahs, Laura Cooper, Chad 
Harvey, Helen Hull-Sanders, Ian Kaplan, Kevin Rice, and Joel Tindle made my time in 
the Eubanks lab productive and fun.  I especially thank my good friend, David Broussard, 
for several years of laughter, camaraderie, and stimulating, if not always intellectual 
(thankfully), conversation.  I am deeply grateful for the unwavering love and support of 
my parents, John and DeDe Styrsky, to whom I dedicate this dissertation.  So, too, am I 
thankful for the love and support of my siblings and my extended family.  It has been 
much easier to keep up the pace with so many cheering me on.  Finally, I profoundly 
thank my wife, Jennifer, for her unfailing love, her gentle but persistent encouragement, 
her kindness, and her patience.  Jennifer helped me see this through from start to finish 
and kept me sane and happy all the while. 
 
vii 
Style manual of journal used:  Chapter 1, Biological Control; Chapter 2, 
Proceedings of the Royal Society of London, B; Chapter 3, Ecology; Chapter 4, 
Ecological Applications         
Computer Software used:   Microsoft? Office XP, SAS for Windows?,     
version 9.1, SigmaPlot? version 9.0                            
 
viii 
TABLE OF CONTENTS 
 
LIST OF TABLES ........................................................................................................ix 
LIST OF FIGURES .......................................................................................................xi 
CHAPTER 1: PLANT TRICHOMES INDIRECTLY ENHANCE  
 TRITROPHIC INTERACTIONS INVOLVING A  
 GENERALIST PREDATOR, THE RED IMPORTED 
 FIRE ANT ....................................................................................1 
CHAPTER 2:  ECOLOGICAL CONSEQUENCES OF ANT-HEMIPTERAN 
MUTUALISMS..........................................................................37 
CHAPTER 3:  A FACULTATIVE MUTUALISM BETWEEN COTTON 
APHIDS AND AN INVASIVE ANT INDIRECTLY  
 BENEFITS PLANT REPRODUCTION .....................................72 
CHAPTER 4:  WHEN INVASIVE SPECIES MEET:  PREDICTING THE 
ECOLOGICAL CONSEQUENCES OF A MUTUALISM 
BETWEEN SOYBEAN APHIDS AND RED IMPORTED     
FIRE ANTS..............................................................................115 
 
 
ix 
LIST OF TABLES 
 
CHAPTER 1 
Table 1.  Results of a MANOVA of trichome density on stems, petioles, and  
leaves  of three isolines of soybeans: glabrous, moderately pubescent, and  
densely   pubescent ..................................................................................................29 
Table 2.  Results of split-plot, repeated measures ANCOVA of the effects  
of fire ant density (Amdro treatment) and trichome density (soybean isoline)  
on the abundances of three groups of herbivores on soybean plants......................... 30 
Table 3.  Results of split-plot, repeated measures ANCOVA of the effects  
of fire ant density (Amdro treatment) and trichome density (soybean isoline)  
on the abundances of three groups of natural enemies on soybean plants................. 31 
CHAPTER 2 
Table 1.  Summary of 25 studies that explicitly test for effects on plants  
of mutualistic interactions between ants and honeydew-producing Hemipterans ..... 65 
CHAPTER 3 
Table 1.  Results of repeated measures ANOVA of the effects of cotton aphid  
density and caterpillar density treatments on the mean abundance of fire ants  
and beet armyworm caterpillars on cotton plants, and mean percent leaf damage  
and boll damage by caterpillars in the 2004 field cage experiment .........................103 
 
x 
Table 2.  Results of repeated measures ANOVA of the effects of cotton aphid  
density and caterpillar density treatments on the mean number of squares,  
flowers, bolls, and nodes on cotton plants in the 2004 field cage experiment......... 104 
Table 3.  Results of ANOVA of the effects of cotton aphid density and         
caterpillar density treatments on the mean number of bolls and seeds, and mean 
seedcotton mass harvested from cotton plants in the 2004 field cage experiment....105 
CHAPTER 4 
Table 1.  Results of repeated measures ANCOVA of the effects of artificial  
honeydew on A. predator abundance on soybean plants and B. herbivore     
abundance and caterpillar damage to leaves in field experiment 1. ........................ 143 
Table 2.  Results of repeated measures ANCOVA of the effects of artificial  
honeydew on taxa diversity of herbivores, predators, and all arthropods on      
soybean plants in field experiment 1. .................................................................... 144 
Table 3.  Results of repeated measures ANCOVA of the effects of artificial  
honeydew on the abundance of fire ants, other predators, and caterpillars, and on 
caterpillar damage to leaves of individual soybean plants in field experiment 2..... 145 
Table 4.  Results of repeated measures ANOVA on the effects of fire ants and 
soybean aphids on the survival of beet armyworm caterpillars on soybean plants       
in greenhouse experiment 1................................................................................... 146 
 
xi 
LIST OF FIGURES 
 
CHAPTER 1 
Figure 1.  Mean (? 1 SE) number of fire ants per beat-cloth sample of soybean  
foliage in high-fire-ant (untreated) and low-fire-ant (Amdro-treated) plots                 
of three soybean isolines varying in trichome density in 2002 ................................. 32 
Figure 2.  Mean (? 1 SE) number of caterpillars per beat-cloth sample of soybean 
foliage over the 2002 growing season in high-fire-ant and low-fire-ant plots  
of three soybean isolines varying in trichome density.............................................. 33 
Figure 3.  Mean (? 1 SE) number of (a) leafhoppers and (b) three-cornered alfalfa 
hoppers per beat-cloth sample of soybean foliage in high-fire-ant and low-fire ant 
plots of three soybean isolines varying in trichome density in 2002......................... 34 
Figure 4.  Mean (? 1 SE) number of (a) big-eyed bugs and (b) damsel bugs per  
beat-cloth sample of soybean foliage over the 2002 growing season in  
high-fire-ant and low-fire-ant plots of soybean plants.............................................. 35 
Figure 5.  Mean (? 1 SE) percent leaf damage by caterpillars over the 2002  
growing season in high-fire-ant and low-fire-ant plots of three soybean isolines 
varying in trichome density..................................................................................... 36 
 
xii 
CHAPTER 2 
Figure 1.  Potential interactions involving a mutualism between ants and  
honeydew-producing Hemipterans.......................................................................... 70 
Figure 2.  Conditionality in the consequences to a host plant of an ant- 
Hemipteran mutualism based on the intensity of non-Hemipteran herbivory ........... 71 
CHAPTER 3 
Figure 1.  Effect of cotton aphid density (mean # per leaf) on fire ant abundance  
on cotton plants (A), effect of fire ant abundance on mean percent leaf damage         
by beet armyworm caterpillars (B), and effect of cotton aphid density on number      
of cotton flowers (C), number of cotton bolls (D), boll mass (E), and leaf and        
stem mass (F) in the 2003 field cage experiment ................................................... 107 
Figure 2.  Effect of the cotton aphid manipulation (aphids included or excluded  
from cotton plants in field cages) on mean cotton aphid density (mean number  
per leaf)  (A), mean fire ant abundance per plant (B), and mean beet armyworm 
caterpillar abundance per plant (C) on seven sampling dates in the 2004 field 
cage experiment.................................................................................................... 109 
Figure 3.  Effect of cotton aphid density (aphids included or excluded from  
cotton plants in field cages) on mean number of cotton squares (flower buds) 
per plant at low and high beet armyworm caterpillar density averaged over seven 
sampling dates (A), and effect of cotton aphid density and caterpillar density on  
mean number of cotton bolls undamaged by caterpillars per plant on seven 
sampling dates (B) in the 2004 field cage experiment............................................ 111 
 
 
xiii 
Figure 4.  Effect of cotton aphid density (aphids included or excluded from  
cotton plants in field cages) on mean number of cotton bolls per plant at harvest 
at low and high beet armyworm caterpillar density (A), and on mean number of 
cotton seeds per plant (B) and mean seedcotton mass per plant (C) at harvest  
averaged across the two caterpillar density treatments in the 2004 field cage 
experiment............................................................................................................ 113 
Figure 5.  Mean percent beet armyworm caterpillar survival over 10 days  
on cotton plants with and without cotton aphids in the 2004 greenhouse  
experiment............................................................................................................ 114 
CHAPTER 4 
Figure 1.  Mean (? 1 SE) number of spiders per beat-cloth sample of soybean    
foliage misted with artificial honeydew or water (control treatment) in 2004......... 147 
Figure 2.  (A) Mean (? 1 SE) number of caterpillars per beat-cloth sample of   
soybean foliage, and (B) mean (? 1 SE) percent leaf damage by caterpillars on 
soybean plants misted with artificial honeydew or water (control treatment)               
in 2004.................................................................................................................. 148 
Figure 3.  Mean (? 1 SE) number of (A) fire ants per plant, (B) non-ant           
predators per plant, and (C) caterpillars per plant, and (D) mean (? 1 SE)          
percent leaf damage by caterpillars on individual soybean plants in three      
treatments in 2004: plants misted with water (control), plants misted with        
artificial honeydew, and plants misted with honeydew from which fire ants           
were excluded with Tanglefoot. ............................................................................ 150 
 
xiv 
Figure 4.  Mean (? 1 SE) percent survival of beet armyworm caterpillars over        
four days on soybean plants with and without soybean aphids either exposed             
or not exposed to fire ants in the 2004 greenhouse experiment. ............................. 151 
Figure 5.  Mean (? 1 SE) number of fire ants per plant and mean (? 1 SE)         
percent survival of (A) big-eyed bug nymphs and (B) damsel bug nymphs on 
individual soybean plants in three treatments in the 2005 greenhouse        
experiments: unmanipulated (control) plants, plants misted with artificial    
honeydew, and plants hosting soybean aphids....................................................... 152 
 
 
1 
CHAPTER ONE: 
PLANT TRICHOMES INDIRECTLY ENHANCE TRITROPHIC 
INTERACTIONS INVOLVING A GENERALIST PREDATOR, 
THE RED IMPORTED FIRE ANT 
 
Abstract 
Morphological defense traits of plants such as trichomes potentially compromise 
biological control in agroecosystems because they may hinder predation by natural 
enemies.  To investigate whether plant trichomes hinder red imported fire ants, 
Solenopsis invicta Buren  (Hymenoptera: Formicidae), as biological control agents in 
soybean, field and greenhouse experiments were conducted in which I manipulated fire 
ant density in plots of three soybean isolines varying in trichome density.  Resulting 
treatment effects on the abundance of herbivores, other natural enemies, plant herbivory, 
and yield were assessed.  Trichomes did not inhibit fire ants from foraging on plants in 
the field or in the greenhouse, and fire ant predation of herbivores in the field was 
actually greater on pubescent plants relative to glabrous plants.  Consequently, fire ants 
more strongly reduced plant damage by herbivores on pubescent plants.  This effect, 
however, did not translate into greater yield from pubescent plants at high fire ant 
densities.  Intraguild predation by fire ants, in contrast, was weak, inconsistent, and did 
not vary with trichome density.  Rather than hindering fire ant predation, therefore,  
 
2 
soybean trichomes instead increased fire ant predation of herbivores resulting in 
enhanced tritrophic effects of fire ants on pubescent plants.  This effect was likely the 
result of a functional response by fire ants to the greater abundance of caterpillar prey on 
pubescent plants.  Given the ubiquity of lepidopteran herbivores and the functional 
response to prey shown by many generalist arthropod predators, a positive indirect effect 
of trichomes on predation by natural enemies might be far more common than is 
currently appreciated. 
 
1. Introduction 
Plants employ a battery of physical and chemical defenses against insect 
herbivores including structural traits such as surface waxes, pubescence, and spines, and 
secondary metabolites that act as repellents, toxins, and digestibility-reducers (Marquis, 
1992; Panda and Khush, 1995).  Parasitoids and predators that use herbivores as hosts or 
prey provide an additional, indirect line of plant defense (Price et al., 1980).  Plant 
resistance and parasitism/predation by natural enemies are not necessarily compatible 
means of plant defense, however, because plant resistance traits may directly or indirectly 
inhibit natural enemies (Price et al., 1980; Bottrell et al., 1998; Cortesero et al., 2000; 
Kennedy, 2003).  Surface waxes on leaves, for example, can reduce the searching 
efficiency of natural enemies by decreasing their ability to grip the plant (Eigenbrode and 
Espelie, 1995), and plant allelochemicals ingested by herbivores may reduce their quality 
as hosts or prey (Turlings and Benrey, 1998).  The potential antagonism between host 
plant resistance traits and natural enemies is particularly important in agriculture because 
 
3 
resistance bred into crop plants may preclude biological control (Bottrell et al., 1998; 
Cortesero et al., 2000). 
 Trichomes are hair-like structural elements of the epidermis of plants that may 
confer physical resistance to insect herbivores by impeding locomotion, feeding, and 
oviposition (Levin, 1973).  Glandular trichomes may confer chemical resistance as well 
by exuding noxious secondary plant metabolites (Levin, 1973).  Besides deterring 
herbivores, however, trichomes may also inhibit natural enemies, principally by 
restricting their mobility on plants (Belcher and Thurston, 1982, Treacy et al., 1986, 
1987; Sutterlin and van Lenteren, 1997; Krips et al., 1999; Lovinger et al., 2000).  As a 
result, trichomes decrease herbivore suppression by both parasitoids (Treacy et al., 1986; 
Kaufman and Kennedy, 1989; McAuslane et al., 1995) and predators (Treacy et al., 1985, 
1987; Barbour et al., 1993; Krips et al., 1999).  In lab experiments with green lacewings, 
Chrysopa carnea Stephens (Neuroptera: Chrysopidae), for example, trichomes slowed 
the walking speed of lacewing larvae on tomato plants (Lycopersicon spp.), thereby 
reducing the rate at which the larvae captured herbivorous mites (Fordyce and Agrawal, 
2001).  Similarly, trichomes on poinsettia plants (Euphorbia pulcherrima) reduced 
twelve-spotted lady beetle, Coleomegilla maculata Timberlake (Coleoptera: 
Coccinellidae), predation of whitefly, Trialeurodes vaporariorum (Westwood) 
(Hemiptera: Aleyrodidae), eggs by physically obstructing 4
th
-instar larvae and adults 
(Lucas et al., 2004). 
Despite consistent support for the hypothesis that trichomes hinder generalist 
predators in several lab and greenhouse experiments, consonant evidence under field 
conditions is limited to one of only two field studies.  Treacy et al. (1985) reported an 
 
4 
inverse relationship between trichome density and predation of Helicoverpa zea (Boddie) 
(Lepidoptera: Noctuidae) eggs by the lacewing predator Chrysopa rufilabris (Burmeister) 
in cotton.  In contrast, Obrycki et al. (1983) found that aphid predation by Coccinellid 
and Chrysopid predators was unrelated to trichome density in potato (Solanum 
tuberosum).  Clearly, a need exists for additional studies that investigate the effects of 
trichomes on predation by natural enemies in the field.  To make such studies more useful 
to biological control, however, they should also consider the tritrophic consequences of 
interactions between trichomes and natural enemies to plant damage and plant 
reproduction. 
In this study, I ask whether trichomes hinder red imported fire ants, Solenopsis 
invicta Buren (Hymenoptera: Formicidae), as predators of herbivores and other natural 
enemies in soybean (Glycine max).  Red imported fire ants (?fire ants? from here forward) 
were unintentionally introduced into Alabama approximately 75 years ago and have 
spread throughout the Southeastern United States (Vinson, 1997).  Because they are 
broadly omnivorous, extremely aggressive, and often superabundant, fire ants negatively 
impact invertebrate communities in both natural and managed ecosystems (Vinson, 1997; 
Wojcik et al., 2001; Holway et al., 2002).  Although fire ants are reportedly significant 
predators of insect pests in several crops (reviewed in Taber, 2000; Holway et al., 2002), 
the utility of fire ants as biological control agents has not been well established because 
they may attack or disrupt other predators and parasitoids (Vinson and Scarborough, 
1989, 1991; Tedders et al., 1990; Eubanks et al., 2002; Kaplan and Eubanks, 2002).  
Further, recent work has shown that the magnitude of the effects of fire ants as predators 
varies among crops.  In a comparison of the effects of fire ants on herbivores and natural 
 
5 
enemies in cotton and soybean, for example, Eubanks (2001) reported that fire ants 
negatively affected a greater number of taxa in cotton despite very similar arthropod 
communities and roughly equivalent densities of fire ants in both crops. 
One hypothesis to account for the disparity in the effects of fire ants between 
cotton and soybean is that dense trichomes on soybean stems and leaves hinder the ability 
of fire ants to forage on soybean plants.  Here, I present the results of field and 
greenhouse experiments in which I manipulated fire ant density in plots of three soybean 
isolines that differed in trichome density to test the hypothesis that trichomes hinder fire 
ant predation of herbivores and intraguild predation of other natural enemies in soybean.  
I predicted that fire ants would be more abundant on glabrous plants relative to 
moderately and densely pubescent plants, resulting in greater suppression of herbivores 
and other natural enemies on glabrous plants at high fire ant densities.  Consequently, I 
predicted that glabrous plants would be less damaged by herbivores at high fire ant 
densities, therefore producing a higher yield than pubescent plants. 
 
2. Methods 
2.1. Field experiment 
I conducted the field experiment at the E. V. Smith Research Center of the 
Auburn University Agricultural Experiment Station in Macon County, Alabama, U.S.A. 
from May through October 2002.  In this experiment, fire ant density (low and high) was 
manipulated in field plots of three soybean isolines varying in trichome density using a 2 
x 3 split-plot, randomized block design.  Seeds from three isolines of ?Lee? soybeans 
(provided by the USDA Crop Genetics and Production Research Unit, Stoneville, 
 
6 
Mississippi, U.S.A.) were planted in 76-cm row spacings at 30 seeds / m.  The three 
isolines were isogenic except for trichome density, which ranged from glabrous to 
moderately pubescent to densely pubescent.  Four 6.1-m rows of each isoline were 
planted in four 6.1 x 9.1-m plots arranged as quarters of a larger, 15.2 x 21.3-m field.  Six 
fields were planted in this configuration to result in 24 replicate plots including each of 
the three isolines.  The location of each isoline within each plot was determined at 
random.  Three-m alleys separated adjacent plots within each field, and all six fields were 
separated by at least 150 m. 
Fire ant densities were naturally high at the study area; therefore, I suppressed fire 
ant densities in three of the six study fields by broadcasting approximately 40 g of 
Amdro
?
 per field every other week throughout the growing season.  Amdro is an ant-
specific bait consisting of corn grit impregnated with a soybean oil attractant and 
hydramethylnon, a toxicant.  Amdro is particularly effective at reducing fire ant densities 
because it can eradicate entire colonies by killing queens.  Because fire ants can forage 
several tens of meters from their colonies, however, the smallest area in which fire ant 
densities could be manipulated using Amdro was at the level of individual fields. 
 I sampled soybean plants for arthropods using a beat cloth (Kogan and Pitrie, 
1980) approximately once weekly from beginning bloom in mid-July through beginning 
seed maturity in late September.  A 1-m
2
 cloth was spread between two adjacent rows of 
plants that were then beat vigorously to dislodge canopy arthropods.  All arthropods that 
fell on the cloth were quickly identified and counted.  I then measured the height of one 
haphazardly chosen plant at that same location and quantified leaf damage by herbivores 
on that plant (sampling dates 1, 3 and 6 only) by summing estimates of the percentage 
 
7 
leaf area consumed on ten haphazardly chosen leaves from throughout the canopy and 
dividing by 10 (Kogan and Turnipseed, 1980).  Each isoline in each plot was randomly 
sampled twice on each sampling date (n = 7 dates).  Means of the two samples were log 
(n +1)-transformed for statistical analysis.  Soybeans were mechanically harvested in late 
October to determine yield (g / m
2
) for each isoline in each plot. 
Because Amdro was applied to entire fields, the experimental unit for the fire ant 
treatment was the field (n = 3 replications).  Accordingly, I tested for effects of the 
Amdro treatment and trichome density (isoline) on fire ant abundance on plants using a 
split-plot, repeated-measures ANCOVA in which Amdro treatment was the whole-plot 
factor and trichome density was the sub-plot factor (SAS proc mixed; SAS Institute, Inc, 
2001).  The same model was used to test for the effects of fire ant abundance (Amdro 
treatment) and trichome density on herbivore abundance, natural enemy abundance, and 
leaf damage by herbivores.  Because plant height varied among plots and fields, plant 
height was included as a covariate in these models to control for associated variation in 
arthropod abundance.  I tested for the effects of fire ant abundance and trichome density 
on soybean yield using a split-plot ANCOVA with plant height as a covariate.  
Experiment-wise error rate was controlled in post-hoc pairwise comparisons of treatment 
means using the Tukey-Kramer adjustment.  Means ? 1 SE are presented in the text and 
P-values ? 0.05 are considered significant in all analyses.  Based on my predictions, I 
was most interested in statistical interactions between the effects of fire ant abundance 
and trichome density.  For brevity and clarity, therefore, only the results of relevant F-
tests in the analyses of treatment effects are presented in the text. 
 
8 
To verify differences in trichome densities among the three isolines, I haphazardly 
collected six individual plants of each isoline from the field plots in early August.  
Trichome density on each plant was estimated by counting the total number of trichomes 
on a 1-cm length of stem and petiole, and on a 1-cm
2
 section of leaf using a dissecting 
microscope with an ocular micrometer.  Trichome density (number / cm
2
) on stems and 
petioles was calculated using the formula for the surface area of a cylinder (? x diameter 
x length).  Stems were sampled between the 6
th
 and 7
th
 node, and trichomes were counted 
on the petiole and underside of one fully expanded leaf haphazardly chosen from the 
upper half of the plant.  I tested for differences in trichome densities among the three 
isolines using a MANOVA (proc glm; SAS Institute, Inc., 2001).  Means ? 1 SE are 
presented in the text. 
 
2.2. Greenhouse experiment 
I conducted the greenhouse experiment on the campus of Auburn University, 
Auburn, Alabama, U.S.A. in July and August 2002.  In this experiment, the presence and 
absence of both fire ants and big-eyed bug nymphs, Geocoris punctipes Say (Hemiptera: 
Geocoridae), was manipulated on plants of three soybean isolines that varied in trichome 
density using a 2 x 2 x 3 randomized complete block design.  The experiment was 
designed to test 1) whether trichomes disrupt fire ant predation of an important soybean 
herbivore, the velvetbean caterpillar, Anticarsia gemmatalis (H?bner) (Lepidoptera: 
Noctuidae), 2) whether trichomes disrupt big-eyed bug predation of velvetbean 
caterpillars, and 3) whether trichomes disrupt fire ant intraguild predation of big-eyed 
bugs.  I used big-eyed bugs in this experiment because they are one of the most abundant 
 
9 
generalist predators in soybean and they readily attack small caterpillars (personal 
observation). 
Seeds from the three ?Lee? isolines described above were sown in 20.3-cm pots 
and the seedlings (n = 40 of each isoline) were allowed to grow until they developed their 
fourth set of true leaves.  Individual potted plants were placed in cages constructed of a 
PVC pipe frame (30 x 33 x 75 cm) that sat within 37-L plastic tubs filled to the rim of the 
pot with potting soil.  Mosquito netting covered each frame and was fastened around the 
outside rim of each tub with an elastic band.  Tubs with plants that were randomly 
assigned to the ?fire ants present? treatment were connected by a short section of 5-cm 
diameter plastic tubing to plastic pans (53 x 43 x 13 cm) containing a small colony of 
approximately 2000 fire ants.  The plastic pans were lined with Fluon? to prevent fire 
ants from escaping.  Fire ants were allowed to access the cages for 24 h prior to the 
initiation of the experiment in order to give them time to discover the plants. 
At the initiation of the experiment, seven 2
nd
-instar velvetbean caterpillars were 
placed on each plant in each treatment.  At the same time, three 4
th
- or 5
th
-instar big-eyed 
bug nymphs were placed in those cages with plants randomly assigned to the ?big-eyed 
bugs present? treatment.  Velvetbean caterpillars were lab-reared on artificial diet 
(Product #F9219B, Bio-Serv, Frenchtown, NJ) from commercially purchased eggs.  Big-
eyed bug nymphs were collected from agricultural fields on the Auburn University 
campus, transferred to an environmental chamber, and provided only water for 48 h prior 
to the experiment.  Densities of fire ants, big-eyed bugs, and velvetbean caterpillars used 
in the experiment are all within the range of densities commonly encountered in Alabama 
soybean fields (Eubanks, 2001). 
 
10 
I attempted to replicate each treatment combination 10 times but fire ants failed to 
forage in three cages.  These cages, therefore, were reassigned to the ?fire ants absent? 
treatment.  Twenty-four hours after initiating the experiment, I recorded the number of 
fire ants, big-eyed bug nymphs, and caterpillars on the plants.  Missing big-eyed bug 
nymphs and caterpillars were considered depredated (Eubanks et al., 2002).  I analyzed 
the effects of fire ants, big-eyed bugs, and trichome density (isoline) on caterpillar 
survival (n surviving / 7) using a three-way ANOVA (SAS proc glm; SAS Institute, Inc., 
2001).  I analyzed the effects of fire ants and trichome density on big-eyed bug survival 
(n surviving / 3) using a two-way ANOVA.  Raw survival data were used in both 
analyses because residuals were normally distributed and homoscedastic.  Error 
probabilities estimated from type-III sums of squares were used to account for the 
unbalanced design.  Experiment-wise error rate was controlled in post-hoc pairwise 
comparisons of treatment means using the Tukey-Kramer adjustment.  Means ? 1 SE are 
presented in the text and P-values ? 0.05 were considered significant in both analyses. 
 
3. Results 
3.1. Field experiment 
3.1.1. Trichome density 
Trichome density varied significantly among the three soybean isolines (Table 1).  
Trichome density was significantly greater in both the moderately pubescent and densely 
pubescent isoline compared to the glabrous isoline (which had no trichomes), and was 
significantly greater in the densely pubescent isoline (stem = 1370 ? 128 trichomes / cm
2
; 
petiole = 663 ? 70; leaves = 823 ? 128) than in the moderately pubescent isoline       
 
11 
(stem = 477 ? 55; petiole = 335 ? 40; leaves = 350 ? 35) (Table 1).  Standardized 
canonical coefficients indicate that the differences among the isolines were due primarily 
to differences in the density of trichomes on the stems, followed by the petioles, and then 
the leaves (Table 1). 
 
3.1.2. Fire ants 
The effects of the Amdro treatment and trichome density interacted to influence 
the abundance of fire ants on plants (F
2,457
 = 3.74, P = 0.03).  Averaged across all 
sampling dates, fire ant abundance was statistically indistinguishable from zero in each 
soybean isoline in Amdro-treated (?low-fire-ant?) plots (Fig. 1).  Although more fire ants 
were sampled in untreated (?high-fire-ant?) plots, fire ant abundance on plants was still 
low, averaging between two and four fire ants per two row-meters (i.e., per beat-cloth 
sample) (Fig. 1).  In high-fire-ant plots, fire ant abundance on plants varied with trichome 
density such that fire ants were slightly more abundant on moderately pubescent plants 
than on densely pubescent plants (P = 0.01, Fig. 1).  Mean fire ant abundance was no 
greater, however, on glabrous plants than on pubescent plants in high-fire-ant plots (Fig. 
1). 
 
3.1.3. Herbivores 
The three most commonly sampled herbivore taxa (accounting for 94% of the 
total number of individuals sampled) were Noctuid larvae (including velvetbean, soybean 
looper [Pseudoplusia includens (Walker)], green cloverworm [Plathypena scabra 
(Fabricius)], and armyworm [Spodoptera sp.] caterpillars), leafhoppers (Hemiptera: 
 
12 
Cicadellidae; including potato leafhoppers [Empoasca fabae (Harris)]), and three-
cornered alfalfa hoppers, Spissistilus festinus (Say) (Hemiptera: Membracidae).  The 
effects of fire ant abundance and trichome density interacted to influence the abundance 
of caterpillars on plants.  Caterpillar abundance differed significantly among the three 
isolines over the season (isoline x date interaction, Table 2) such that caterpillars were 
more abundant on pubescent plants than on glabrous plants (Fig. 2).  The effect of fire ant 
abundance varied among isolines, however, such that fire ants reduced caterpillar 
abundance more in the two pubescent isolines than in the glabrous isoline as the season 
progressed (fire ant x isoline x date interaction, Table 2; Fig. 2).  Averaged over the final 
three sampling dates (when caterpillar abundance peaked), fire ants reduced caterpillar 
abundance by 34.5%, 27.5%, and 3.1% on moderately pubescent, densely pubescent, and 
glabrous plants, respectively. 
Fire ant abundance and trichome density also interacted to affect leafhopper 
abundance on plants (fire ant x isoline interaction, Table 2).  Unlike caterpillars, 
leafhoppers were significantly more abundant on glabrous plants than on pubescent 
plants.  Similar to their effect on caterpillars, however, fire ants more greatly reduced 
leafhopper abundance in the two pubescent isolines (Fig. 3a).  Fire ants reduced mean 
leafhopper abundance by 75.1%, 75.7%, and 5.1% on moderately pubescent, densely 
pubescent, and glabrous plants, respectively. 
Fire ants had the opposite effect on three-cornered alfalfa hoppers.  Three-
cornered alfalfa hoppers were significantly more abundant over the season in high-fire-
ant plots than in low-fire-ant plots (fire ant x date interaction, Table 2).  The beneficial 
effect of fire ants on S. festinus abundance varied with trichome density (fire ant x isoline 
 
13 
interaction, Table 2), however, such that alfalfa hoppers were more abundant on 
moderately pubescent plants than on densely pubescent plants at high fire ant densities (P 
< 0.0001; Fig. 3b).  Alfalfa hopper abundance was no greater on glabrous plants relative 
to pubescent plants (Fig. 3b). 
 
3.1.4. Natural enemies 
The three most commonly sampled natural enemy taxa (accounting for 92% of the 
total number of individuals sampled) were big-eyed bugs (Geocoridae), damsel bugs 
(Nabidae), and spiders (Araneae).  Fire ants negatively affected big-eyed bugs and 
damsel bugs, but did so independently of trichome density.  Significantly fewer big-eyed 
bugs (Fig. 4a) and damsel bugs (Fig. 4b) were found on plants in high-fire-ant plots than 
in low-fire-ant plots, but only on single sampling dates (fire ant x date interactions, Table 
3).  Trichome density also affected big-eyed bug and damsel bug abundances on plants 
(isoline effects, Table 3).  Averaged over all sampling dates, big-eyed bugs were more 
abundant on moderately pubescent (3.8 ? 0.3 bugs/sample) and densely pubescent plants 
(3.0 ? 0.3) than on glabrous plants (1.9 ? 0.3; P < 0.0001 for both comparisons).  In 
contrast, damsel bug abundance did not differ between pubescent plants and glabrous 
plants (1.5 ? 0.5 bugs/sample); however, damsel bugs were more abundant on moderately 
pubescent plants (1.7 ? 0.5) relative to densely pubescent plants (1.3 ? 0.5; P = 0.003).  
In contrast to big-eyed bugs and damsel bugs, neither fire ant abundance nor trichome 
density affected spider abundance on plants (Table 3). 
 
 
 
14 
3.1.5. Plant damage and yield 
Caterpillars were likely responsible for the majority of leaf herbivory as they were 
the most abundant chewing herbivores sampled.  As a consequence of the effects of fire 
ant abundance and trichome density on caterpillar abundance, the effects of fire ant 
abundance and trichome density interacted to influence the level of caterpillar damage to 
plants.  Percent leaf damage differed significantly among the three isolines over the 
season (isoline x date interaction: F
4,193
 = 23.81, P < 0.0001) such that pubescent plants 
were more damaged by caterpillars than were glabrous plants (Fig. 5).  The effect of fire 
ants varied among isolines, however, such that pubescent plants rather than glabrous 
plants were significantly less damaged by caterpillars over the season in high-fire-ant 
plots relative to low-fire-ant plots (fire ant x isoline x date interaction: F
4,193
 = 2.83, P = 
0.03; Fig. 5).  On the final date that plants were scored for damage, caterpillar herbivory 
of moderately pubescent and densely pubescent plants was 13% and 10% lower, 
respectively, in high-fire-ant plots compared to low-fire-ant plots (Fig. 5).  Conversely, 
caterpillar herbivory of glabrous plants was 4% higher in high-fire-ant plots compared to 
low-fire-ant plots (Fig. 5). 
Although the effects of fire ant abundance and trichome density interacted to 
influence caterpillar herbivory, the effect of the fire ant x isoline interaction on yield was 
not statistically significant (F
2,62
 = 0.21, P = 0.81).  Soybean yield did vary significantly 
among isolines (F
2,62
 = 3.67, P = 0.03); however, despite the fact that glabrous plants 
were less damaged by herbivores, yield from glabrous plants (5.9 ? 0.4 g/m
2
) was no 
greater than that of either pubescent isoline.  Yield differed statistically only between 
moderately pubescent plants (5.7 ? 0.3 g/m
2
) and densely pubescent plants                   
 
15 
(4.8 ? 0.3 g/m
2
; P = 0.03).  In contrast, fire ant abundance on plants did not significantly 
affect yield (F
1,4
 = 1.05, P = 0.36), despite a 27% increase in yield in high-fire-ant plots 
(6.1 ? 0.9 g/m
2
) relative to low-fire-ant plots (4.8 ? 0.9 g/m
2
). 
 
3.2. Greenhouse experiment 
In cages with fire ants, trichome density did not affect the number of fire ants 
foraging on plants (glabrous: 1.1 ? 0.3 ants/plant; moderately pubescent: 0.9 ? 0.3; 
densely pubescent: 1.2 ? 0.3; one-way ANOVA: F
2,54
 = 0.31, P = 0.74).  Likewise, 
trichome density did not affect the ability of fire ants to depredate velvetbean caterpillars 
(fire ant x isoline interaction: F
2,105
 = 0.31, P = 0.73).  Significantly fewer caterpillars 
survived for 24 h in the presence of fire ants (68.7 ? 2.3%) than in the absence of fire ants 
(93.5 ? 2.2%), regardless of trichome density (F
1,105 
= 63.95, P < 0.0001).  In contrast, 
caterpillar survival was not affected by either big-eyed bug nymphs (F
1,105 
= 1.79, P = 
0.18) or trichome density (F
2,105 
= 0.34, P = 0.71). 
Although fire ants reduced caterpillar survival, fire ants had no effect on the 
survival of big-eyed bug nymphs (F
1,50  
= 1.89, P = 0.18).  Trichome density, however, 
did have a marginally significant effect on nymph survival (F
2,50 
= 3.09, P = 0.054).  
Nymph survival was greater on moderately pubescent plants (64.7 ? 7.0%) relative to 
glabrous plants (41.3 ? 7.0%; P = 0.05).  Nymph survival did not differ between densely 
pubescent plants (59.3 ? 7.1%) and either glabrous or moderately pubescent plants. 
 
 
 
16 
4. Discussion 
4.1. Soybean trichomes enhance fire ant predation of herbivores 
Plant trichomes are generally thought to impede arthropod predator-prey 
interactions by disrupting predator mobility and decreasing searching efficiency (Bottrell 
et al., 1998; Cortesero et al., 2000).  Despite supporting evidence from several lab and 
greenhouse experiments, however, little corroborating evidence exists that trichomes 
have the same effect in the field.  In addition to a general lack of field data, no previous 
studies have shown whether the effects of trichomes on predation of herbivores have any 
consequence to plant damage or plant reproduction.  Finally, no studies to date have 
investigated the effects of trichomes on ants, a very common and important group of 
generalist predators that may prey on other natural enemies as well as herbivores 
(H?lldobler and Wilson, 1990; Holway et al., 2002). 
In this study, I conducted field and greenhouse experiments to test the hypothesis 
that trichomes hinder predation by red imported fire ants in soybean resulting in 
increased plant damage by herbivores and decreased yield.  My results suggest that rather 
than hindering fire ant predation, soybean trichomes instead increased fire ant predation 
of herbivores resulting in decreased plant damage.  Firstly, fire ants were no more 
abundant on glabrous plants than on moderately or densely pubescent plants either in the 
field or in the greenhouse.  In fact, fire ants were slightly more numerous on pubescent 
plants in the field.  Secondly, increased trichome density either had no effect or even a 
positive effect on fire ant predation of arthropods.  Fire ant predation of caterpillars did 
not vary with trichome density in the greenhouse, whereas fire ant predation of 
herbivores including several caterpillar species in the field was generally greater on 
 
17 
pubescent plants than on glabrous plants.  Further, intraguild predation by fire ants, 
though inconsistent and weak in the field, was no greater on glabrous plants than on 
pubescent plants.  Thirdly, fire ants more strongly suppressed caterpillar herbivory on 
pubescent plants than on glabrous plants in the field.  This effect did not result in greater 
yield from pubescent plants because plant damage was relatively greater on pubescent 
plants, but, averaged over all three isolines, yield was almost 30% greater in high-fire-ant 
plots.  Combined, these results demonstrate a trophic cascade whereby the direct effect of 
a natural enemy on an herbivore population indirectly benefits plant fitness, at least in 
terms of reduced plant damage (Schmitz et al., 2000).  In this particular tritrophic 
interaction, however, the cascade is apparently contingent upon the expression of a 
morphological resistance trait by the host plant.  I am not aware of any previous report in 
the literature of such a phenomenon. 
Our results are inconsistent with most previous studies investigating the effect of 
trichomes on generalist predators other than ants.  Trichomes reportedly decrease the 
foraging efficiency of coccinellid larvae and adults (Belcher and Thurston, 1982; 
Obryicki and Tauber, 1984; Lucas et al., 2004), lacewing larvae and adults (Treacy et al., 
1985, 1987; Fordyce and Agrawal, 2001), an Anthocorid bug (Coll et al., 1997), a 
predaceous stink bug (De Clercq et al., 2000), and a predaceous mite (Krips et al., 1999) 
by impeding predator movement and prey encounter rate.  The effect of trichomes on 
intraguild predation has been far less studied, but in one laboratory experiment, Lucas 
and Brodeur (1999) found that dense pubescence on potato plants reduced the 
susceptibility of the eggs of a predatory midge to predation by the lady beetle C. 
maculata.  All but one of these studies (Treacy et al., 1985) was conducted under 
 
18 
controlled conditions in labs and greenhouses, and none focused on ants.  In the 
comparison of a greenhouse and a field study of aphid predation, Obrycki and Tauber 
(1984) suggested that coccinellid and chrysopid predators were less hindered by 
glandular trichomes on potato plants in the field because the effects of dust, wind, and 
rain decreased trichome adhesiveness.  Soybean trichomes are predominantly             
non-glandular, however, and do not produce sticky exudates.  Perhaps trichomes simply 
do not compromise the mobility of ants on plants as they do for other generalist 
predators.  I did not compare fire ant prey encounter rates among isolines, but the fact 
that fire ants suppressed caterpillars more on pubescent plants than on glabrous plants in 
the field without a concomitant increase in ant abundance on pubescent plants argues 
against the possibility that trichomes hindered fire ant searching efficiency. 
I hypothesize that trichomes increased fire ant predation of soybean herbivores as 
an indirect consequence of their effect on the abundance of herbivore prey.  Soybean 
trichomes provide strong resistance to leafhoppers (Turnipseed, 1977, this study [Fig. 
3a]) but reportedly facilitate oviposition by lepidopteran herbivores such as the corn 
earworm (Lambert and Kilen, 1989), velvetbean caterpillar (Gregory, 1989), and soybean 
looper (Beach and Todd, 1988).  Such an oviposition preference by gravid moths may 
explain why caterpillars were more abundant on pubescent plants in both high-fire-ant 
and low-fire-ant plots (Fig. 2).  Increased fire ant predation of caterpillars on pubescent 
plants relative to glabrous plants (Fig. 2), therefore, may represent a functional response 
by fire ants to the greater abundance of caterpillar prey on pubescent plants.  Indeed, 
density-dependent predation is a common attribute of generalist predators including ants 
(Risch and Carroll, 1982; Way and Khoo, 1992; Harvey and Eubanks, 2004).  The fact 
 
19 
that trichomes did not increase fire ant predation of caterpillars in the greenhouse 
experiment supports this interpretation.  In contrast to the field experiment, caterpillar 
abundance was standardized across the three soybean isolines in the greenhouse 
experiment, thus precluding a functional response by fire ants. 
 
 
4.2. Fire ants as biological control agents in soybean 
Previous work has shown that fire ants can be both useful and disruptive as 
predators of cotton arthropods (Eubanks 2001, Eubanks et al. 2002, Kaplan and Eubanks 
2002, 2005).  Fire ants can reduce the abundance of important cotton pests such as Lygus 
plant bugs (Hemiptera: Miridae) and caterpillars, for example, but can also increase the 
abundance of another important cotton pest, the cotton aphid, Aphis gossypii Glover 
(Hemiptera: Aphididae), by attacking natural enemies of the aphid.  Given that trichomes 
increased fire ant predation of an important group of soybean herbivores but did not 
appreciably increase intraguild predation in this study, my results highlight the role of fire 
ants as important beneficial predators in soybean.  Fire ants can reach extremely high 
densities in soybean fields in the southeastern United States, with mound densities 
ranging from 22.2 to 207.5 active mounds per ha (Banks et al., 1990).  Although fire ants 
may be the most abundant natural enemies in soybean fields, few studies have 
investigated their effects as predators of herbivores and no studies have investigated 
intraguild predation by fire ants in the soybean canopy.  Fire ants reportedly prey on 
velvetbean caterpillar eggs (Buschman et al., 1977), larvae (Elvin et al., 1983), and pupae 
(Lee et al., 1990) in soybean, as well as on the eggs of the southern green stink bug, 
 
20 
Nezara viridula (Linnaeus) (Hemiptera: Pentatomidae), an important pod-feeding pest 
(Krispyn and Todd, 1982).  My results provide even broader evidence that fire ants can 
significantly reduce the abundance of herbivore pests in soybean and that fire ant 
suppression of defoliating herbivores potentially benefits soybean yield.   
Unfortunately, the very traits that make fire ants potentially useful in biological 
control, namely their abundance, aggression, and omnivorous feeding habit, may also 
make them a threat to yield.  Fire ants may reduce soybean yield directly, for example, by 
feeding on germinating seeds, seedlings, and the root nodules of growing soybean plants 
(Shatters and Vander Meer, 2000).  Further, fire ants may indirectly harm soybean plants 
by tending S. festinus nymphs for their honeydew, thereby protecting them from other 
predators.  Spissitilus festinus nymphs and adults can cause extensive yield losses by 
feeding on petioles and racemes during the pod-setting and pod-filling stages (Kogan and 
Turnipseed, 1987).  I found that S. festinus nymphs and adults were more abundant over 
the season on plants in high-fire-ant plots than in low-fire-ant plots, suggesting that fire 
ants disrupted biological control of this herbivore.  Whether fire ants directly or indirectly 
harmed plants in this study was apparently of little consequence, however, because yield 
did not differ statistically between low-fire-ant and high-fire-ant plots. 
 
4.3. Conclusion 
In summary, results from field and greenhouse experiments show that soybean 
trichomes do not inhibit fire ants from foraging on plants but rather increase fire ant 
predation of herbivores, possibly as an indirect consequence of greater caterpillar 
abundance on pubescent plants.  I suggest that this phenomenon might actually be quite 
 
21 
common in nature because lepidopteran larvae are ubiquitous herbivores, many generalist 
arthropod predators including ants show density-dependent predation responses to prey 
density, and ants are often abundant predators in both natural and managed systems 
(H?lldober and Wilson, 1990; Way and Khoo, 1992; Holway et al., 2002).  I emphasize 
the need for additional field studies of the effects of trichomes on generalist predators and 
their consequences to biological control.  As demonstrated in many lab and greenhouse 
experiments, trichomes may reduce the mobility and prey encounter rate of some 
predator species.  Numerical or functional responses to prey density by these same 
predators in the field, however, may compensate for reduced mobility resulting in 
increased plant protection.  In contrast to previous studies, therefore, my results suggest 
that trichomes, though providing resistance to some herbivores (e.g., leafhoppers) but not 
to others (e.g., lepidopteran herbivores) in soybean, are compatible with predation by a 
generalist predator as means of direct and indirect plant defense. 
 
Acknowledgments 
Thomas Kilen of the USDA-ARS Crop Genetics and Production Research Unit provided 
seeds of the three isolines of ?Lee? soybeans.  James Bannon, Steve Nightengale, and the 
staff of the E. V. Smith Research Center?s Plant Breeding Unit provided exceptional 
logistical support.  I thank Andrew Borin, Sarah Carter, Lydia Daniel, and Elizabeth May 
for assistance with the fieldwork.  Financial support was provided by the Alabama Fire 
Ant Management Program and the Department of Entomology and Plant Pathology, 
Auburn University. 
 
 
22 
References 
Banks, W.A., Adams, C.T, Lofgren, C.S., Wojcik., D.P., 1990. Imported fire ant 
infestation of soybean fields in the southern United States. Florida Entomol. 73, 
503-504. 
Barbour, J.D., Farrar, Jr., R.R., Kennedy, G.G., 1993. Interaction of Manduca sexta 
resistance in tomato with insect predators of Helicoverpa zea. Entomol. Exp. App. 
68, 143-155. 
Beach, R.M., Todd, J.W., 1988. Oviposition preference of the soybean looper 
(Lepidoptera: Noctuidae) among four soybean genotypes differing in larval 
resistance. J. Econ. Entomol. 81, 344-348. 
Belcher, D.W., Thurston, R., 1982. Inhibition of movement of larvae of the convergent 
lady beetle by leaf trichomes of tobacco. Environ. Entomol. 11, 91-94. 
Bottrell, D.G., Barbosa, P., Gould, F., 1998.  Manipulating natural enemies by plant 
variety selection and modification: a realistic strategy? Annu. Rev. Entomol. 43, 
347-367. 
Buschman, L.L., Whitcomb, W.H., Hemenway, R.C., Mays, D.L., Ru, N., Leppla, N.C., 
Smittle, B.J., 1977. Predators of velvetbean caterpillar eggs in Florida soybeans. 
Environ. Entomol. 6, 403-407. 
Coll, M., Smith, L.A., Ridgway, R.L., 1997. Effect of plants on the searching efficiency 
of a generalist predator: the importance of predator-prey spatial associations. 
Entomol. Exp. App. 83, 1-10. 
Cortesero, A.M., Stapel, J.O., Lewis, W.J., 2000. Understanding and manipulating plant 
attributes to enhance biological control. Biol. Control 17, 35-49. 
 
23 
De Clercq, P., Mogaghegh, J., Tirry, L., 2000. Effect of host plant on the functional 
response of the predator Podisus nigrispinus (Heteroptera: Pentatomidae). Biol. 
Control 18, 65-70. 
Eigenbrode, S.D., Espelie, K.E., 1995. Effects of plant epicuticular lipids on insect 
herbivores. Annu. Rev. Entomol. 40, 171-194. 
Elvin, M.K., Stimac, J.L., Whitcomb, W.H., 1983. Estimating rates of arthropod 
predation on velvetbean caterpillar larvae in soybeans. Florida Entomol. 66, 319-
330. 
Eubanks, M.D., 2001. Estimates of direct and indirect effects of red imported fire ants on 
biological control in field crops. Biol. Control 21, 35-43. 
Eubanks, M.D., Blackwell, S.A, Parrish, C.J., Delamar, Z.D., Hull-Sanders, H., 2002. 
Intraguild predation of beneficial arthropods by red imported fire ants in cotton. 
Environ. Entomol. 31, 1168-1174. 
Fordyce, J.A., Agrawal, A.A., 2001. The role of plant trichomes and caterpillar group 
size on growth and defense of the pipevine swallowtail Battus philenor. J. Anim. 
Ecol. 70, 997-1005. 
Gregory, B., 1989. Field observations of mating, oviposition, and feeding behavior of 
Anticarsia gemmatalis (Lepidoptera: Noctuidae) in a soybean field. J. Insect 
Behav. 2, 761-774. 
Harvey, C.T., Eubanks, M.D., 2004. Effect of habitat complexity on biological control by 
the red imported fire ant (Hymenoptera: Formicidae) in collards. Biol. Control 29, 
348-358. 
H?lldobler, B., Wilson, E.O., 1990. The Ants. Belknap Press, Cambridge, MA. 
 
24 
Holway, D.A., Lach, L., Suarez, A.V., Tsutsui, N.D., Case, T.J., 2002. The causes and 
consequences of ant invasions. Annu. Rev. Ecol. Syst. 33, 181-233. 
Kaplan, I., Eubanks, M.D., 2002. Disruption of cotton aphid (Homoptera: Aphididae)-
natural enemy dynamics by red imported fire ants (Hymenoptera: Formicidae). 
Environ. Entomol. 31, 1175-1183. 
Kaplan, I., Eubanks, M.D., 2005. Aphids alter the community-wide impact of fire ants. 
Ecology 86, 1640-1649. 
Kaufman, W.C., Kennedy, G.G., 1989. Relationship between trichome density in tomato 
and parasitism of Heliothis spp. (Lepidoptera: Noctuidae) eggs by Trichogramma 
spp. (Hymenoptera: Trichogrammatidae). Environ. Entomol. 18, 698-704. 
Kennedy, G.G., 2003. Tomato, pests, parasitoids, and predators: tritrophic interactions 
involving the Genus Lycopersicon. Annu. Rev. Entomol. 48, 51-72. 
Kogan, M., Pitrie, Jr., H.N., 1980. General sampling methods for above-ground 
populations of soybean arthropods. In: Kogan, M., Herzog, D.C. (Eds.), Sampling 
Methods in Soybean Entomology. Springer-Verlag, New York, NY, pp. 30-60. 
Kogan, M., Turnipseed, S.G., 1980. Soybean growth and assessment of damage by 
arthopods. In: Kogan, M., Herzog, D.C. (Eds.), Sampling Methods in Soybean 
Entomology. Springer-Verlag, New York, NY, pp. 3-29. 
Kogan, M., Turnipseed, S.G., 1987. Ecology and management of soybean arthropods. 
Annu. Rev. Entomol. 32, 507-538. 
Krips, O.E., Kleijn, P.W., Willems, P.E.L., Gols, G.J.Z., Dicke, M., 1999. Leaf hairs 
influence searching efficiency and predation rate of the predatory mite 
Phytoseiulus persimilis (Acari: Phytoseiidae). Exp. Appl. Acarol. 23, 119-131. 
 
25 
Krispyn, J.W., Todd, J.W., 1982. The red imported fire ant as a predator of the southern 
green stinkbug in Georgia. J. Georgia Entomol. Soc. 17, 19-26. 
Lambert, L., Kilen, T.C., 1989. Influence and performance of soybean lines isogenic for 
pubescence type on oviposition preference and egg distribution by corn earworm 
(Lepidoptera: Noctuidae). J. Entomol. Sci. 24, 309-316. 
Lee, J., Johnson, S.J., Wright, V.L., 1990. Quantitative survivorship analysis of the 
velvetbean caterpillar (Lepidoptera: Noctuidae) pupae in soybean fields in 
Louisiana. Environ. Entomol. 19, 978-986. 
Levin, D.A., 1973. The role of trichomes in plant defense. Q. Rev. Biol. 48, 3-15. 
Lovinger, A., Liewehr, D., Lamp, W.O., 2000. Glandular trichomes on alfalfa impede 
searching behavior of the potato leafhopper parasitoid. Biol. Control 18, 187-192. 
Lucas, E., Brodeur, J., 1999. Oviposition site selection by the predatory midge 
Aphidoletes aphidimyza (Diptera: Cecidomyiidae). Environ. Entomol. 28, 622-
627. 
Lucas, E., Labrecque, C., Coderre, D., 2004. Delphastus catalinae and Coleomegilla 
maculata lengi (Coleoptera: Coccinellidae) as biological control agents of the 
greenhouse whitefly, Trialeurodes vaporariorum (Homoptera: Aleyrodidae). Pest 
Manage. Sci. 60, 1073-1078. 
Marquis, R.J., 1992. Selective impact of herbivores. In: Fritz, R.S., Simms, E.L. (Eds.), 
Plant Resistance to Herbivores and Pathogens: Ecology, Evolution, and Genetics. 
Univ. Chicago Press, Chicago, IL, pp. 301-325. 
 
26 
McAuslane, H.J., Johnson, F.A., Colvin, D.L., Sojack, B., 1995. Influence of foliar 
pubescence on abundance and parasitism of Bemesia argentifolia (Homoptera: 
Aleyrodidae) on soybean and peanut. Environ. Entomol. 24, 1135-1143. 
Obrycki, J.J., Tauber, M.J., 1984. Natural enemy activity on glandular pubescent potato 
plants in the greenhouse: an unreliable predictor of effects in the field. Environ. 
Entomol. 13, 679-683. 
Obrycki, J.J., Tauber, M.J., Tingey, W.M., 1983. Predator and parasitoid interaction with 
aphid-resistant potatoes to reduce aphid densities: a two-year field study. J. Econ. 
Entomol. 74, 456-462. 
Panda, N., Khush, G.S., 1995. Host Plant Resistance to Insects. CAB International, 
Oxford. 
Price, P.W., Bouton, C.E., Gross, P., McPheron, B.A., Thompson, J.N., Weis, A.E., 1980. 
Interactions among three trophic levels: influence of plants on interactions 
between insect herbivores and natural enemies. Annu. Rev. Ecol. Syst. 11, 41-65. 
Risch, S.J., Carroll, C.R., 1982. The ecological role of ants in two Mexican 
agroecosystems. Oecologia 55, 114-119. 
SAS Institute, Inc., 2001. SAS for Windows, Version 8.2. SAS Institute Inc., Cary, NC. 
Schmitz, O.J., Hamback, P.A., Beckerman, A.P., 2000. Trophic cascades in terrestrial 
systems: a review of the effects of carnivore removals in plants. Am. Nat. 155, 
141-153. 
Shatters, R.G., Vander Meer, R.K., 2000. Characterizing the interaction between fire ants 
(Hymenoptera: Formicidae) and developing soybean plants. J. Econ. Entomol. 93, 
1680-1687. 
 
27 
Sutterlin, S., van Lenteren, J.C., 1997. Influence of hairiness of Gerbera jamesonii leaves 
on the searching efficiency of the parasitoid Encarsia formosa. Biol. Control 9, 
157-165. 
Taber, S.W., 2000. Fire Ants. Texas A & M Univ. Press, College Station, TX. 
Tedders, W.L., Reilly, C.C., Wood, B.W., Morrison, R.K., Lofgren, C.S., 1990. Behavior 
of Solenopsis invicta (Hymneoptera: Formicidae) in pecan orchards. Environ. 
Entomol. 19, 44-53. 
Treacy, M.F., Benedict, J.H., Lopez, J.D., Morrison, R.K., 1987. Functional response of a 
predator (Neuroptera: Chrysopidae) to bollworm (Lepidoptera: Noctuidae) eggs 
on smoothleaf, hirsute, and pilose cottons. J. Econ. Entomol. 80, 376-379. 
Treacy, M.F., Benedict, J.H., Segers, J.C., Morrison, R.K., Lopez, J.D., 1986. Role of 
cotton trichome density in bollworm (Lepidoptera: Noctuidae) egg parasitism. 
Environ. Entomol. 15, 365-368. 
Treacy, M.F., Zummo, G.R., Benedict, J.H., 1985. Interactions of host-plant resistance in 
cotton with predators and parasites. Agric. Ecosyst. Environ. 13, 151-157. 
Turlings, T.C.J., Benrey, B., 1998. Effects of plant metabolites on the behavior and 
development of parasitic wasps. Ecoscience 5, 321-333. 
Turnipseed, S.G., 1977. Influence of trichome variations on populations of small 
phytophagous insects in soybean. Environ. Entomol. 6, 815-817. 
Vinson, S.B., 1997. Invasion of the red imported fire ant (Hymenoptera: Formicidae): 
spread, biology, and impact. Am. Entomol. 43, 23-39. 
 
28 
Vinson, S.B., Scarborough, T.A., 1989. Impact of the imported fire ant on laboratory 
populations of cotton aphid (Aphis gossypii) predators. Florida Entomol. 72, 107-
11. 
Vinson, S.B., Scarborough, T.A., 1991. Interactions between Solenopsis invicta 
(Hymenoptera: Formicidae), Rhopalosiphum maidis (Homoptera: Aphididae), and 
the parasitoid ? Lysiphlebus testaciepes Cresson (Hymenoptera: Aphidiidae). 
Ann. Entomol. Soc. Am. 84, 158-164. 
Way, M.J., Khoo, K.C., 1992. Role of ants in pest management. Annu. Rev. Entomol. 37, 
479-503. 
Wojcik, D.P., Allen, C.R., Brenner, R.J., Forys, E.A., Jouvenaz, D.P, Lutz, R.S., 2001. 
Red imported fire ants: impact on biodiversity. Am. Entomol. 47, 16-23. 
  
29 
Ta
b
l
e 1
.
  R
e
s
u
l
t
s
 o
f
 a
 M
A
N
O
V
A
 o
f
 t
r
i
c
h
o
m
e
 d
e
n
s
i
t
y
 o
n
 s
t
e
m
s
,
 p
e
t
i
o
l
e
s
, a
n
d
 
l
e
a
v
e
s
 o
f
 t
h
re
e
 
i
s
o
l
i
n
e
s
 o
f
 
s
o
y
b
ea
n
s
: g
l
a
b
r
o
us
,
 m
o
der
a
t
e
l
y
 pu
b
e
s
c
e
n
t
,
 an
d den
s
e
l
y
 pu
b
e
s
c
e
n
t
.
 
  
 
 
 
 
S
t
an
dar
d
i
zed 
ca
n
o
n
.
 
c
o
e
f
f
i
ci
en
t
a
 
 
 Wi
l
k
s
?
 
?
 
F 
d.
f
.
 
P 
St
e
m
 Pe
t
i
o
l
e
 
L
e
a
f
 
M
a
i
n
 
e
f
f
ect
 
 
 
  
  
 
  
  
 I
s
o
l
i
n
e 
0.
0419 
16.
84 
6,
 26 
<0.
0001 
2.
527 
1.
238 
0.
713 
Co
n
t
r
a
s
t
s 
 
  
 
  
 
  
  
 Gl
a
b
r
o
us
 v
s
.
 M
o
der
a
t
e
 
0.
2319 
14.
36 
3,
 13 
  0
.
0002 
2.
171 
1.
706 
0.
520 
  
  
 Gl
a
b
r
o
us
 v
s
.
 Den
s
e 
0.
0485 
85.
07 
3,
 13 
<0.
0001 
2.
511 
1.
263 
0.
704 
  
  
 M
o
der
a
t
e
 v
s
.
 Den
s
e 
0.
1237 
30.
69 
3,
 13 
<0.
0001 
2.
696 
0.
935 
0.
816 
a 
Va
l
u
e
s
 r
e
pr
es
e
n
t
 es
t
i
m
a
t
e
s
 o
f
 t
h
e 
t
o
t
a
l
 a
m
o
u
n
t
 
o
f
 v
a
r
i
at
i
o
n
 
i
n
 t
r
i
c
h
o
m
e
 de
n
s
i
t
y
 acco
un
t
e
d f
o
r
 by
 ea
c
h
 
v
a
r
i
a
b
l
e
 a
l
o
n
e.
  T
h
e l
a
r
g
es
t
 v
a
l
u
e w
i
t
h
i
n
 a s
p
ec
i
f
i
c
 
t
e
s
t
 i
n
d
i
cat
es
 t
h
e v
a
r
i
a
b
l
e
 t
h
at
 m
o
s
t
 di
f
f
er
ed 
i
n
 t
r
i
c
h
o
m
e
 
d
e
n
s
i
t
y
 be
t
w
e
e
n
 is
o
lin
e
s
.
 
 
  
30 
Ta
b
l
e 2
.
  Res
u
l
t
s
 o
f
 s
p
l
i
t
-
p
l
o
t
,
 
r
e
peat
ed
 m
e
as
ur
es
 
A
N
COV
A
 o
f
 t
h
e e
f
f
ect
s
 o
f
 
f
i
r
e
 a
n
t
 den
s
i
t
y
 (
A
m
d
r
o
 
t
r
ea
t
m
e
n
t
)
 an
d t
r
i
c
h
o
m
e
 
d
e
n
s
i
t
y
 (
s
o
y
b
e
an
 
i
s
o
l
i
n
e
)
 on
 th
e
 a
b
un
d
a
n
c
es
 of
 th
r
e
e
 g
r
ou
p
s
 
of
 
h
e
r
b
i
v
or
es
 on
 s
o
y
b
e
an
 pl
an
ts
.
 
  
C
a
t
e
rp
i
l
l
a
rs
 
 
L
e
a
f
h
o
p
p
e
rs
 
 
A
l
f
a
l
f
a
 
h
o
p
p
e
rs
 
F
act
o
r
 d.
f
.
 
F P 
F 
P 
F 
P 
F
i
r
e
 a
n
t
s
 
1
,
4
 
  
  
  
  1
.
9
0
 
  
  
  0
.
2
4
 
  
  
  
  4
.
8
7
 
  
  
  0
.
0
9
 
  
  
  
 1
.
8
8
 
  
  
 0
.
2
4
 
I
s
o
l
i
n
e 
2,
457 
  
  
  
  5
.
22 
  
  
  0
.
006 
  
  115
.
19 
  
  <0.
0001 
  
  
 13.
83 
  
 <0.
0001 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
457 
  
  
  71
.
27 
  
  <0.
0001 
  
  
  
  3
.
98 
  
  
  0
.
05 
  
  
  
 3.
70 
  
  
 0.
06 
Sa
m
p
li
ng dat
e
 
6,
457 
  
  243
.
87 
  
  <0.
0001 
  
  
  51
.
70 
  
  <0.
0001 
  
  
  
 6.
87 
  
 <0.
0001 
Fi
r
e
 an
t
s
 x
 Dat
e
 
6,
457 
  
  
  
  3
.
69 
  
  
  0
.
001 
  
  
  
  3
.
20 
  
  
  0
.
004 
  
  
  
 6.
79 
  
  
 0.
0001 
I
s
o
l
i
n
e 
x
 Dat
e
 
12,
457 
  
  
  
  5
.
30 
  
  <0.
0001 
  
  
  
  4
.
13 
  
  <0.
0001 
  
  
  
 1.
50 
  
  
 0.
12 
Fi
r
e
 an
t
s
 x
 I
s
o
l
i
n
e 
2,
457 
  
  
  
  0
.
18 
  
  
  0
.
83 
  
  
  
  4
.
96 
  
  
  0
.
007 
  
  
  
 3.
32 
  
  
 0.
04 
Fi
r
e
 an
t
s
 x
 I
s
o
l
i
n
e x
 Dat
e
 
12,
457 
  
  
  
  3
.
51 
  
  <0.
0001 
  
  
  
  0
.
64 
  
  
  0
.
81 
  
  
  
 0.
58 
  
  
 0.
86 
 
  
31 
Ta
b
l
e 3
.
  Res
u
l
t
s
 o
f
 s
p
l
i
t
-
p
l
o
t
,
 
r
e
peat
ed
 m
e
as
ur
es
 
A
N
COV
A
 o
f
 t
h
e e
f
f
ect
s
 o
f
 
f
i
r
e
 a
n
t
 den
s
i
t
y
 (
A
m
d
r
o
 
t
r
ea
t
m
e
n
t
)
 an
d t
r
i
c
h
o
m
e
 
d
e
n
s
i
t
y
 (
s
o
y
b
e
an
 
i
s
o
l
i
n
e
)
 on
 th
e
 a
b
un
d
a
n
c
es
 of
 th
r
e
e
 g
r
ou
p
s
 
of
 
n
a
tu
r
a
l
 en
em
i
e
s
 on
 s
o
y
b
e
an
 pl
an
ts
.
 
 
 
B
i
g-
e
y
ed b
ugs
 
 
Da
m
s
e
l
 b
u
g
s
 
 
Sp
i
d
er
s
 
F
act
o
r
 d.
f
.
 
F 
P 
F P F 
P 
F
i
r
e
 a
n
t
s
 
1
,
4
 
  
  
  
 0
.
1
7
 
  
  
  0
.
7
0
 
  
  
  
  0
.
2
3
 
  
  
  0
.
6
5
 
  
  
  
 0
.
7
2
 
  
  
  0
.
4
4
 
I
s
o
l
i
n
e 
2,
457 
  
  
 35.
72 
  
  <0.
0001 
  
  
  
  5
.
37 
  
  
  0
.
005 
  
  
  
 2.
47 
  
  
  0
.
09 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
457 
  
  
  
 4.
77 
  
  
  0
.
03 
  
  
  
  0
.
46 
  
  
  0
.
50 
  
  
  
 9.
58 
  
  
  0
.
002 
Sa
m
p
li
ng dat
e
 
6,
457 
  
  
 38.
02 
  
  <0.
0001 
  
  
  51
.
22 
  
  <0.
0001 
  
  
  
 6.
59 
  
  <0.
0001 
Fi
r
e
 an
t
s
 x
 Dat
e
 
6,
457 
  
  
  
 3.
05 
  
  
  0
.
006 
  
  
  
  5
.
22 
  
  <0.
0001 
  
  
  
 0.
76 
  
  
  0
.
60 
I
s
o
l
i
n
e 
x
 Dat
e
 
12,
457
  
  
  
 2.
46 
  
  
  0
.
004 
  
  
  
  1
.
32 
  
  
  0
.
20 
  
  
  
 0.
79 
  
  
  0
.
66 
Fi
r
e
 an
t
s
 x
 I
s
o
l
i
n
e 
2,
457 
  
  
  
 0.
73 
  
  
  0
.
48 
  
  
  
  0
.
88 
  
  
  0
.
41 
  
  
  
 0.
18 
  
  
  0
.
84 
Fi
r
e
 an
t
s
 x
 I
s
o
l
i
n
e x
 Dat
e
 
12,
457
  
  
  
 0.
71 
  
  
  0
.
75 
  
  
  
  0
.
92 
  
  
  0
.
53 
  
  
  
 1.
23 
  
  
  0
.
26 
 
 
32 
Fig. 1.  Mean (? 1 SE) number of fire ants per beat-cloth sample of soybean foliage in 
high-fire-ant (untreated) and low-fire-ant (Amdro-treated) plots of three soybean isolines 
varying in trichome density in 2002.  The application of Amdro effectively reduced the 
abundance of fire ants on plants and fire ant abundance varied only slightly among 
isolines in high-fire-ant plots. 
Soybean isoline
Fi
r
e
 
ant
s
 /
 sam
pl
e
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
high-fire-ant plots
low-fire-ant plots
Glabrous Moderately
pubescent
Densely
pubescent
 
 
33 
Fig. 2.  Mean (? 1 SE) number of caterpillars per beat-cloth sample of soybean foliage 
over the 2002 growing season in high-fire-ant and low-fire-ant plots of three soybean 
isolines varying in trichome density.  Caterpillars were least abundant on glabrous plants 
later in the season but fire ants most reduced caterpillar abundance on pubescent plants. 
 
Low-fire-ant plots High-fire-ant plots
Julian date
C
a
ter
p
illa
r
s
 / s
a
m
p
le
0
5
10
15
20
25
30
35
Glabrous
Moderately pubescent
Densely pubescent
208 222 236 250 264 208 222 236 250 264
 
 
34 
Fig. 3.  Mean (? 1 SE) number of (a) leafhoppers and (b) three-cornered alfalfa hoppers 
per beat-cloth sample of soybean foliage in high-fire-ant and low-fire ant plots of three 
soybean isolines varying in trichome density in 2002.  Fire ants most reduced leafhopper 
abundance on pubescent plants.  Fire ants positively affected alfalfa hopper abundance 
only in the moderately pubescent isoline. 
A
l
fa
lfa h
o
p
p
er
s / 
sam
p
l
e
0.00
0.25
0.50
0.75
1.00
Le
afho
pp
er
s / sam
pl
e
0
2
4
6
8
high-fire-ant plots
low-fire-ant plots
(a)
(b)
Glabrous Moderately
pubescent
Densely
pubescent
Soybean Variety
 
 
35 
Fig. 4.  Mean (? 1 SE) number of (a) big-eyed bugs and (b) damsel bugs per beat-cloth 
sample of soybean foliage over the 2002 growing season in high-fire-ant and low-fire-ant 
plots of soybean plants.  Natural enemies were significantly more abundant in the low-
fire-ant plots on the dates indicated with asterisks. 
D
a
m
s
e
l
 
bug
s /
 sa
mp
l
e
0
1
2
3
4
(b)
B
i
g-
e
y
e
d bug
s / 
s
a
m
p
le
0
1
2
3
4
5
6
high-fire-ant plots
low-fire-ant plots
(a)
*
*
Julian date
208 222 236 250 264
 
 
36 
Fig. 5.  Mean (? 1 SE) percent leaf damage by caterpillars over the 2002 growing season 
in high-fire-ant and low-fire-ant plots of three soybean isolines varying in trichome 
density.  Fire ants suppressed caterpillar herbivory only on pubescent plants. 
Julian date
%
 cat
er
pi
l
l
ar
 
h
e
r
b
i
v
o
r
y
0
10
20
30
40
50
60
Glabrous
Moderately pubescent
Densely pubescent
208 222 236 250 264 208 222 236 250 264
Low-fire-ant plots High-fire-ant plots
 
 
37 
CHAPTER TWO: 
ECOLOGICAL CONSEQUENCES OF ANT-HEMIPTERAN MUTUALISMS 
 
Mutualisms between ants and honeydew-producing Hemipterans are abundant and 
widespread in arthropod food webs, yet their ecological consequences are very poorly 
known.  Ant-Hemipteran mutualisms have potentially broad ecological effects because 
the presence of honeydew-producing Hemipterans dramatically alters the abundance and 
predatory behavior of ants on plants.  I review several studies that investigate explicitly 
the consequences of ant-Hemipteran mutualisms as ?keystone interactions? to arthropod 
communities and their host plants.  Ant-Hemipteran mutualisms have mostly negative 
effects on the local abundance and species richness of several guilds of herbivores and 
predators.  In contrast, of 25 studies that document the effects of ant-Hemipteran 
mutualisms on plants, the majority (68%) show that plants actually benefit from these 
mutualisms.  In these cases, increased predation or harassment of other, more damaging 
herbivores by Hemipteran-tending ants resulted in decreased plant damage by other 
herbivores.  Both the community-level and host plant effects of ant-Hemipteran 
mutualisms are likely conditional, however, depending on specific ecological contexts.  
The ecological consequences of mutualisms between Hemipterans and invasive ants 
relative to native ants have rarely been studied but may be particularly important because 
of the greater abundance, aggressiveness, and extreme omnivory of invasive ants.  I argue 
 
38 
that ant-Hemipteran mutualisms are largely overlooked and underappreciated 
interspecific interactions that have strong and pervasive effects on the communities in 
which they are embedded. 
 
1. Introduction 
Mutualism, defined as a reciprocally beneficial interaction between individuals of two 
species, is increasingly recognized as a common and important ecological interaction 
(Boucher et al. 1982, Bronstein 1994a, Stachowicz 2001).  Perhaps one of the most 
familiar examples is the food-for-protection mutualism between ants (Hymenoptera: 
Formicidae) and honeydew-producing Hemipteran insects.  Many ant species ?tend? 
Hemipteran insects in the suborders Sternorrhyncha (specifically aphids, whiteflies, 
scales, and mealybugs) and Auchenorrhyncha (specifically treehoppers and leafhoppers) 
for their honeydew (H?lldobler & Wilson 1990).  Honeydew is a sugary excretion of 
carbohydrates, amino acids, and water derived from plant phloem upon which 
Hemipterans feed (Way 1963).  Ants are attracted to honeydew as a predictable, 
renewable food resource and consequently protect Hemipterans from predators and 
parasitoids (Way 1963, Buckley 1987). 
Ant-Hemipteran mutualisms are very common, if not well-appreciated, 
interactions in terrestrial communities from temperate to tropical latitudes (Buckley 1987, 
H?lldobler & Wilson 1990, Delabie 2001).  Ants tend honeydew-producing Hemipterans 
on an extremely wide range of plants, including grasses (e.g., Moya-Ragoza & Nault 
2000), forbs (e.g., Messina 1981, Renault et al. 2005), and vines, shrubs, and trees (e.g., 
Floate & Whitham 1994, Rico-Gray & Castro 1996, Way et al. 1999).  In a study of the 
 
39 
ant community in the Amazonian rainforest canopy, Bl?thgen et al. (2000) recorded ant-
Hemipteran associations on 20 of 24 tree genera and 41 of 66 individual trees surveyed.  
Mutualistic interactions between ants and Hemipterans are also very common in managed 
(e.g., agricultural) habitats in addition to being widespread in natural habitats from 
grasslands to forests (Buckley 1987, Way & Khoo 1992). 
The vast majority of studies of mutualisms between ants and honeydew-producing 
Hemipterans have focused historically on the costs and benefits to both partners.  In 
contrast, ant-Hemipteran mutualisms have received comparatively little scrutiny from a 
broader, community-level perspective, particularly their potential effects on trophic 
interactions in food webs, arthropod community structure, and host plant fitness.  
Predatory species of ants are important constituents of arthropod food webs because of 
their active and catholic feeding habits, efficiency in recruiting to and subduing prey, and 
sheer abundance (H?lldobler & Wilson 1990).  Mutualistic interactions between ants and 
honeydew-producing Hemipterans may dramatically increase the effects of ants as 
predators, however, because the presence of Hemipterans alters ant behavior such that 
ants become more aggressive and attack insects that they might otherwise ignore, 
including other predators as well as other herbivores (Way 1963) (Figure 1).  Further, 
Hemipteran honeydew supports entire colonies of arboreal ants, allowing them to 
maintain much greater densities than could be supported by scavenging and predation 
alone (Davidson 1997, Bl?thgen et al. 2000, Davidson et al. 2003), and attracts ground-
dwelling ants onto plant foliage, increasing the likelihood that they encounter arthropods 
in plant-based food webs (Way & Khoo 1992, Del-Claro & Oliveira 1999, Kenne et al. 
2003, Kaplan & Eubanks 2005). 
 
40 
Because the strength of interactions between ants and other species in the local 
arthropod community may be enhanced in the presence of honeydew-producing 
Hemipterans, Hemipteran-tending ants may also indirectly affect plant fitness.  Increased 
ant predation of other herbivores may benefit plants if those herbivores impose a greater 
cost to plant growth and reproduction than do Hemipterans (Buckley 1987, Way & Khoo 
1992, Delabie 2001).  In contrast, by protecting Hemipterans from predators and 
parasitoids, and by stimulating Hemipteran feeding rate, fecundity, and dispersal, tending 
ants may exacerbate the negative effects of Hemipterans on plants, including stunted 
growth, reduced leaf area, and the introduction of plant pathogens, all of which can 
decrease plant fitness (Way 1963, Beattie 1985, Buckley 1987, Delabie 2001).  The 
consequences of ant-Hemipteran mutualisms to plant fitness, thus, should represent a 
trade-off between the indirect benefit of increased ant suppression of other herbivores and 
the direct cost of herbivory by Hemipterans (Carroll & Janzen 1973, Buckley 1987, Lach 
2003) (Figure 1). 
Here, I review the published literature on the ecological consequences of ant-
Hemipteran mutualisms.  My goal is to highlight mutualistic interactions between ants 
and honeydew-producing Hemipterans as ?keystone interactions? that have broad yet 
underappreciated effects on the abundance and distribution of species in arthropod 
communities and on host plant fitness in both natural and managed systems.  Further, I 
discuss evidence that the consequences of ant-Hemipteran mutualisms vary in different 
ecological contexts.  I also discuss the extreme paucity of information regarding the 
ecological consequences of mutualistic interactions between Hemipterans and invasive 
ants.  I conclude by considering several reasons why ecologists have overlooked the 
 
41 
ecological consequences of ant-Hemipteran mutualisms and I argue that these mutualisms 
are important components of terrestrial food webs that may dramatically influence 
community structure and trophic interactions. 
 
2. Effects on community structure and species diversity 
 The vast majority of previous studies of the effects of ant-Hemipteran mutualisms 
on other arthropods have focused primarily on ant predation or harassment of predators 
and parasitoids of Hemipterans.  That Hemipteran-tending ants reduce the survival, 
abundance, and alter the spatial distribution of the natural enemies of Hemipterans on 
plants is extensively documented (e.g., Bradley 1973, Cushman and Whitham 1989, 
Tedders et al. 1990, Stechmann et al. 1996, Del-Claro & Oliveira 2000, Kaplan & 
Eubanks 2002, Renault et al. 2005).  Fewer studies have focused on the effects of ant-
Hemipteran mutualisms on other herbivores sharing the same host plant, but in most 
cases Hemipteran-tending ants significantly reduce their survival and abundance (e.g., 
Nickerson et al. 1977, Messina 1981, Skinner & Whittaker 1981, Ito & Higashi 1991, 
Floate & Whitham 1994, Moya-Raygoza & Nault 2000, Sipura 2002, Suzuki et al. 2004, 
Kaplan & Eubanks 2005). 
Only three studies have investigated the effects of ant-Hemipteran mutualisms on 
the broader community (multiple species of predators and herbivores) in which they are 
embedded.  These studies have shown that mutualistic interactions between ants and 
Hemipterans alter the abundance and distribution of specialist and generalist predators 
and parasitoids, and multiple species of herbivores in several feeding guilds, resulting in 
changes to local species diversity.  In an experimental study of the effects of the ant 
 
42 
Formica lugubris on the guild structure of herbivorous insects in birch (Betula 
pubescens), Fowler and MacGarvin (1985) found that the abundance of the ant-tended 
aphid, Symydobius oblongus, increased by 8200% on branches with ants compared to 
branches from which ants had been excluded.  In contrast, species richness of non-tended 
Hemipterans was reduced by 28%, species richness of leaf-chewing caterpillars was 
reduced by 69%, and total herbivore species richness was reduced by 28% on branches 
with ants.  The results of a survey of herbivores on birches at ant and non-ant sites were 
similar and also showed that species richness of beetles (Coleoptera) was 61% lower on 
trees with aphid-tending ants.  In contrast, species richness of leaf-mining caterpillars, a 
guild of herbivores that were not vulnerable to ant predation, was actually 44% greater on 
trees with aphid-tending ants, presumably because the ants indirectly protected the 
concealed caterpillars from other predators (Fowler & MacGarvin 1985). 
Wimp and Whitham (2001) more directly tested for community-level effects of an 
ant-Hemipteran mutualism by manipulating the presence and absence of the ant-tended 
aphid Chaitophorus populicola in a cottonwood (Populus sp.) system.  On trees from 
which aphids were removed, tending Formica propinqua abandoned the trees resulting in 
a 76% increase in herbivore abundance and a 76% increase in the abundance of generalist 
predators.  Moreover, total arthropod abundance increased by 80% and total species 
richness increased by 57% on trees from which aphids were removed.  Likewise, a 
mutualistic interaction between cotton aphids (Aphis gossypii) and red imported fire ants 
(Solenopsis invicta) strongly affected the abundance and distribution of other arthropods 
in cotton (Gossypium hirsutum) (Eubanks 2001, Eubanks et al. 2002, Kaplan and 
Eubanks 2002, 2005).  In large-scale field manipulations of fire ant densities, the 
 
43 
mutualism between S. invicta and A. gossypii negatively affected the abundance of 27-
33% of herbivore taxa and 40-47% of predator taxa (Kaplan & Eubanks 2005).  Similar 
to Wimp and Whitham (2001), the percentage of taxa that was negatively affected by the 
ant-aphid mutualism did not differ between herbivores and predators (although there was 
a trend towards a stronger effect on predators). 
 These three studies provide compelling evidence that ant-Hemipteran mutualisms 
can act as ?keystone interactions? that dramatically alter the structure of arthropod 
communities on plants.  In the presence of honeydew-producing Hemipterans, tending 
ants disproportionately affected complex food webs by directly or indirectly altering the 
abundance of many interacting species.  Predation or harassment by tending ants 
typically reduced the abundance of other herbivores, generalist predators, and specialist 
predators and parasitoids.  Further, ant-Hemipteran mutualisms typically reduced local 
species diversity.  In some cases, however, ant-Hemipteran mutualisms may actually 
benefit specific herbivore species or herbivore guilds that are not as vulnerable to ant 
predation (e.g., leaf-mining caterpillars) by reducing the abundance of their natural 
enemies (Fowler & MacGarvin 1985; see also Fritz 1983).  Ant-Hemipteran mutualisms 
may even increase arthropod diversity within a plant community if the mutualisms occur 
only on specific plant species (or genotypes) and certain arthropod species (e.g., 
specialist natural enemies of Hemipterans) are found only on those plants hosting the 
mutualism (Wimp & Whitham 2001).  
 
 
 
 
44 
3. Effects on host plant damage and fitness 
Given both the ecological and economic significance of Hemipterans as plant 
pests, and the widespread distribution and abundance of both Hemipterans and ants in 
terrestrial communities, surprisingly few studies have quantified the consequences of ant-
Hemipteran mutualisms to plant fitness.  Table 1 lists 25 studies that have explicitly 
tested for an effect of the interaction between ants and honeydew-producing Hemipterans 
on plants by either 1) comparing plants on which Hemipterans and/or tending ants are 
naturally present or absent, or 2) by comparing plants on which the densities of 
Hemipterans and/or tending ants has been manipulated. 
I note that of the 18 studies that were experimental rather than correlative, 14 
employed ant-exclusion manipulations (Table 1).  Although perhaps easier than 
manipulating Hemipteran densities, a major limitation of ant-exclusion experiments is 
that they do not specifically test the net effect of the interaction per se between ants and 
Hemipterans on plants (i.e., whether ants have stronger effects as predators in the 
presence of Hemipterans).  In many systems, ants forage on plants even in the absence of 
Hemipterans.  Experiments in which Hemipteran density is manipulated (Hemipterans 
either excluded from plants or added to plants where there were previously none) (n = 2 
studies; Table 1) provide a more realistic test of this hypothesis because ant behavior 
changes in response to the presence of honeydew-producing Hemipterans rather than vice 
versa (see Floate & Whitham 1994).  Better still are experiments that manipulate both 
ants and Hemipterans in a crossed design (n = 2 studies; Table 1) because they allow 
comparison of the effects of ants alone and the effects of Hemipterans alone with the 
effects of the interaction between ants and Hemipterans specifically. 
 
45 
 
(a) net negative effects 
Despite abundant evidence that tending ants can cause outbreaks of honeydew-
producing Hemipterans (Beattie 1985, Buckley 1987, Holway et al. 2002) and that 
Hemipteran insects can impose tremendous costs on plant fitness via phloem 
consumption and transmission of plant pathogens (Carter 1962, Buckley 1987, Delabie 
2001), negative effects of mutualistic interactions between ants and Hemipterans on plant 
fitness have rarely been demonstrated (n = 7 studies; Table 1).  Whether negative effects 
of ant-Hemipteran mutualisms are truly uncommon in nature or whether they have simply 
been assumed and, thus, not quantified is unclear.  In the only published study of a 
negative effect in a managed system, Banks & Macaulay (1967) reported that the aphid 
Aphis fabae was 30-50% more numerous on faba bean (Vicia faba) when tended by the 
ant Lasius niger, resulting in a 26% reduction in pods per plant, a 53% reduction in seeds 
per plant, and a 38% reduction in seed mass relative to plants with A. fabae that were not 
tended by ants.  Similarly, a facultative mutualism between the ant Camponotus planatus 
and an unidentified aphid species hosted by the woody vine Paullinia fuscecens reduced 
the proportion of inflorescences with seeds by 7-23% and reduced seed number per 
inflorescence by 46-60% compared to plants with untended aphids (Rico-Gray & Castro 
1996).  In a study of the effects of an ant-aphid mutualism on the fitness of the forb 
Bidens pilosa, Renault et al. (2005) reported that the aphid Aphis coreopsidis was 34% 
more abundant in the presence of tending Camponotus ants than in their absence.  
Consequently, plants hosting ant-tended aphids produced significantly fewer viable seeds 
than did plants from which ants were excluded.  Finally, in the only study of which I am 
 
46 
aware that documents an effect of an ant-Hemipteran mutualism on the transmission of a 
plant pathogen, the presence of S. invicta caused a 240% increase in aphid abundance on 
tomato (Lycopersicon esculentum), resulting in a 65% increase in the level of infection of 
plants by the aphid-vectored Cucumber Mosaic Virus (Cooper 2005). 
 
(b) no net effect 
 Two published studies failed to find any net effect, negative or positive, of ant-
Hemipteran mutualisms on plants (Table 1).  The ant Formica subsericea protected the 
treehopper Vanduzca arquata from its natural enemies on black locust (Robinia 
pseudoacacia), but increased treehopper densities in the presence of ants did not harm 
plants, apparently because the plants were tolerant of treehopper herbivory (Fritz 1983).  
The tending ants also attacked the adults and eggs of a leaf-mining beetle but indirectly 
protected beetle larvae by excluding a damsel bug predator, resulting in no net gain or 
loss in plant growth.  Similarly, Moya-Raygoza & Nault (2000) found that although ants 
protected the leafhopper Dalbulus quinquenotatus from predation by spiders, populations 
of the leafhopper on its gamagrass host plants (Tripsacum pilosum and T. dactyloides) 
never reached damaging levels in the field, apparently as a result of incidental predation 
of leafhoppers by the tending ants.  Further, even though tending ants reduced the 
abundance of two nonmyrmecophilous leafhopper species, the ant-leafhopper mutualism 
provided no net benefit to the plant. 
 
 
 
 
47 
(c) net positive effects 
Studies documenting net positive effects of ant-Hemipteran mutualisms on plants 
are more numerous (n = 17; Table 1).  In these studies, ant-Hemipteran mutualisms 
benefited plants as a consequence of increased ant predation or harassment of other, more 
damaging insects, resulting in decreased plant damage by those insects and/or increased 
plant fitness.  Treehopper-tending ants on Solidago altissima, for example, deterred 
feeding by the larvae of two species of leaf beetles (Messina 1981).  Increased plant 
protection by ants in the presence of treehoppers resulted in increased plant growth and 
seed production relative to neighboring plants on which treehoppers were naturally 
absent.  Similarly, oaks (Quercus dentata) that hosted the aphid Tuberculatus quercicola 
had a significantly lower percentage of acorns infested by acorn-boring Lepidopteran 
larvae in areas where the aphid-tending ant Formica yessensis was abundant relative to 
areas where the ants were rare (Ito & Higashi 1991). 
Skinner and Whittaker (1981) showed that leaf damage by Lepidopteran larvae 
was significantly reduced on sycamore trees (Acer pseudoplatanus) on which the ant 
Formica rufa tended the aphid Periphyllus testudinaceus compared to trees from which 
ants were experimentally excluded.  In later ant-exclusion experiments in the same 
system, Whittaker and Warrington (1985) showed that the radial growth of sycamore 
trees was 2-3 times greater in the presence of the ant-aphid mutualism, presumably as a 
consequence of herbivore suppression by the aphid-tending ants.  Floate and Whitham 
(1994) manipulated aphid densities rather than ant densities to test whether a facultative 
mutualism between the aphid Chaitophorus populicola and a Formica ant species 
indirectly protected cottonwood (Populus fremontii and hybrids) foliage from beetle 
 
48 
herbivory.  The addition of aphids to trees dramatically increased ant abundance, which 
translated into a 95% reduction in the abundance of beetle larvae and a 50% reduction in 
leaf damage. 
Six of the studies that documented a net positive effect were conducted in 
managed systems (Table 1), demonstrating that ant-Hemipteran mutualisms may be 
exploited for biological control of more damaging pests of agriculturally important 
plants.  In fact, arboreal ants attracted to Hemipteran honeydew have long been utilized to 
control non-Hemipteran herbivore pests in several tree crops (Way and Khoo 1992).  
Growers manipulate colonies of Dolichoderus ants in cocoa (Theobroma cacao) and 
sapodilla (Manilkara zapota) plantations in Southeast Asia, for example, because the 
mealybug-tending ants protect cocoa pods and sapodilla fruits from damage by Mirids 
(Heteroptera) (Khoo and Ho 1992) and Pyralid larvae (Lepidoptera) (Van Mele and Cuc 
2001) without any direct yield loss caused by the mealybugs.  Far less is known, 
however, about the net effects of ant-Hemipteran mutualisms on herbaceous crop plants 
in more ephemeral systems such as row and vegetable crops.  Protection by tending ants 
can lead to Hemipteran outbreaks in agricultural monocultures of herbaceous plants 
resulting in significantly reduced yield caused by loss of plant sap (e.g., Banks & 
Macaulay 1967) and transmission of plant pathogens (e.g., Cooper 2005).  If particular 
Hemipteran species do not vector plant pathogens to certain herbaceous crop plants and if 
those plants show high tolerance to Hemipteran herbivory, ant-Hemipteran mutualisms 
may still benefit plant yield.  In cotton, for example, I have shown experimentally that A. 
gossypii attract ground-foraging S. invicta onto plants, resulting in increased ant 
predation of an important cotton herbivore, the beet armyworm caterpillar (Spodoptera 
 
49 
exigua).  Cotton plants hosting the mutualism suffered less caterpillar damage to leaves 
and bolls (cotton fruit) and, consequently, produced 16% more bolls and 10% more 
seedcotton relative to plants from which A. gossypii were excluded (Chapter 3).  I 
encourage additional studies of the effects of ant-Hemipteran mutualisms in other annual 
cropping systems to further evaluate their utility for increased biological control of non-
Hemipteran herbivores. 
 
4.  Conditionality in the effects of ant-Hemipteran mutualisms on plants 
 Conditionality in the outcome of mutualistic interactions between ants and 
Hemipterans and between ants and plants has been widely noted (Bronstein 1994b).  
Because the net effects of ant-Hemipteran mutualisms on plants represent a trade-off 
between the direct cost of feeding by Hemipterans and the indirect benefit of ant 
suppression of other herbivores, any biotic or abiotic factor that influences the magnitude 
of the cost or benefit may alter the consequence of the interaction to the plant (Lach 
2003).  Such conditionality has been often suggested but rarely tested.  Factors that could 
mediate the effects of ant-Hemipteran mutualisms on plant fitness include: 1) the 
proximity of ant colonies to host plants (because distance may affect the numerical 
response of ants to honeydew-producing Hemipterans; Karhu 1998, Wimp & Whitham 
2001), 2) the proximity and abundance of other plants hosting honeydew-producing 
Hemipterans that may distract tending ants (Cushman & Whitham 1991), 3) the 
abundance and efficacy of other arthropod predators, and the effects of tending ants on 
those predators (Fritz 1983), 4) the abundance and relative cost to plant fitness of insects 
in other herbivore guilds (Messina 1981, Fritz 1983, Chapter 3), 5) indirect interactions 
 
50 
between Hemipteran and non-Hemipteran herbivores mediated by host plants (Stout et al. 
1998, Rodriguez-Saona et al. 2005), 6) host plant tolerance of and resistance to other 
herbivores (Sipura 2002; see also Vrieling et al. 1991), 7) variation in plant vulnerability 
to Hemipteran and non-Hemipteran herbivory associated with plant phenology 
(Rosenheim et al. 1997), 8) the availability and quality of plant extrafloral nectar 
(Buckley 1983), 9) the effects of plant genotype on plant quality (Floate & Whitham 
1994), and 10) the effects of other biotic (e.g., pathogen infection) and abiotic factors 
(e.g., nitrogen availability, shade) on host plant quality (Cushman 1991).   
As suggested by Messina (1981) and Fritz (1983), the consequences of an ant-
Hemipteran mutualism to a particular host plant may vary simply with the intensity of 
herbivory by other herbivores on the plant (Figure 2).  I tested this hypothesis in an 
experiment in which I manipulated the presence and absence of the aphid A. gossypii, and 
densities of S. exigua caterpillars (low and high) on cotton plants in a field that was 
naturally infested by S. invicta.  As described earlier, S. invicta workers were more 
abundant on plants with A. gossypii, resulting in increased predation of caterpillars.  The 
fitness benefit of reduced caterpillar damage to plants varied with caterpillar density, 
however, such that plant protection by aphid-tending S. invicta resulted in a 32% increase 
in bolls (relative to plants without aphids) at high caterpillar density versus just a 3% 
increase in bolls at low caterpillar density (Chapter 3).  Thus, the indirect benefit of ant 
suppression of caterpillars substantially outweighed the direct cost of A. gossypii only 
when caterpillars were abundant. 
Only one other study has investigated conditionality in the ecological 
consequences of ant-Hemipteran mutualisms to plants.  Sipura (2002) manipulated the 
 
51 
presence and absence of Formica aquilonia on shoots of two willow species with and 
without aggregations of the aphid Pterocomma salicis to test the effect of the ant-aphid 
mutualism on willow herbivory and growth.  The two willow species differed in 
salicylate concentration in the foliage, a trait that confers resistance to leaf-chewing 
herbivores (Sipura 1999).  Leaf damage was significantly reduced and shoot length and 
radial growth were significantly enhanced in the presence of the mutualism only in the 
willow species with a low concentration of salicylate (Salix phylicifolia) and, thus, low 
resistance to herbivores.  In contrast, the ant-aphid mutualism negatively affected growth 
in the willow species with high resistance (S. myrsinifolia), presumably because the cost 
of aphid herbivory outweighed the very limited benefit of ant suppression of leaf-
chewing insects. 
 
5.  Ecological consequences of ant-Hemipteran mutualisms involving native ants 
versus invasive ants 
The degree of protection provided by tending ants to Hemipterans varies among 
native ant species, primarily because of differences among species in aggressiveness and 
territoriality (Buckley & Gullan 1991, Kaneko 2003, Novgorodova 2005).  Given that 
more aggressive ant species with more strongly defended territories provide greater 
protection to Hemipterans, these ant species may be expected to have stronger direct and 
indirect effects on community structure and host plant fitness.  I am unaware of any 
studies that have compared the effects on arthropod community structure of ant-
Hemipteran mutualisms involving different native ant species, but a few studies suggest 
that more aggressive ant species have stronger indirect effects on plants.  Sipura (2002), 
 
52 
for example, compared the effects of mutualisms between aphids and two native ant 
species on the herbivory and growth of two willow species and found that only the more 
aggressive ant species had measurable effects (both positive and negative, depending on 
the willow species [see section 4]) on plant growth.   
Invasive ants, that subset of unintentionally introduced ant species that establish 
populations in natural areas undisturbed by human activity, are particularly notorious for 
their extreme abundance and aggressiveness, two traits at least partly responsible for their 
pervasive negative ecological effects in invaded habitats (Holway et al. 2002).  Invasive 
ants are also very attracted to Hemipteran honeydew and rapidly recruit in large numbers 
to Hemipteran aggregations from which they may exclude native tending ants (Holway et 
al. 2002, Lach 2003, Ness & Bronstein 2004).  Indeed, exploitation of Hemipteran 
honeydew and plant extrafloral nectar is hypothesized to contribute to the ecological 
dominance of invasive ants (Holway et al. 2002).  Although invasive ants can cause local 
declines in arthropod abundance and diversity (Holway et al. 2002), no studies have 
investigated the direct and indirect effects of ant-Hemipteran mutualisms involving 
invasive ant species on interspecific interactions within arthropod food webs. 
Few data exist as well to evaluate the consequences to plants of ant-Hemipteran 
mutualisms involving invasive ants (Table 1).  Invasive ants are exceptional tenders and 
frequently cause local Hemipteran outbreaks (Holway et al. 2002, Ness & Bronstein 
2004).  Because of their abundance, aggression, and territoriality, however, invasive ants 
may provide comparatively greater protection to plants from other herbivores than do 
native ants (Holway et al. 2002, Lach 2003; but see Rapp & Salum 1995).  Way et al. 
(1999), for example, observed that defoliation of the pine Pinus pinaster by the larvae of 
 
53 
the pine processionary moth (Thaumetopoea pityocampa) was slight to negligible in trees 
patrolled by the invasive Argentine ant (Linepithema humile), but was moderate to severe 
in trees patrolled by several native ant species.  The authors suggested that aggregations 
of honeydew-producing aphids facilitated protection of pines by L. humile, but they did 
not specifically test whether the ants had the same effect in the absence of aphids.  
Facultative mutualisms between the invasive ant S. invicta and aphids in tomato can harm 
plants by increasing the severity of infection by a plant virus (Cooper 2005), but can 
benefit plant reproduction in cotton by increasing ant suppression of more damaging 
herbivores (Chapter 3).  Whether the effects of S. invicta in these systems differ from 
those by native aphid-tending ants is unknown. 
 
6. Why so little information on such a ubiquitous and potentially important 
interaction? 
 Despite their ubiquity in terrestrial communities, mutualistic interactions between 
ants and honeydew-producing Hemipterans have received relatively little attention in 
terms of their effects on food web dynamics and tritrophic interactions involving their 
host plants.  This dearth of information regarding the ecological consequences of ant-
Hemipteran mutualisms potentially represents a major void in our understanding of the 
dynamics of complex trophic interactions in arthropod communities.  My review of the 
literature shows that ant-Hemipteran mutualisms can broadly affect the local abundance 
and distribution of specialist and generalist predators and parasitoids, as well as 
herbivores in several feeding guilds.  In addition, though some evidence shows that     
ant-Hemipteran mutualisms result in a net decrease in the fitness of their host plants, a 
 
54 
majority of studies document an indirect positive effect in which Hemipteran-tending 
ants protect host plants from other, more damaging herbivores.  Because the broad 
ecological effects of Hemipteran-tending ants as predators are precipitated by their 
interaction with Hemipterans (i.e., ant effects would not be as strong in the absence of 
Hemipterans), ant-Hemipteran mutualisms may, thus, represent ?keystone interactions? in 
many communities. 
Why, then, have the ecological consequences of ant-Hemipteran mutualisms been 
so little studied?  I suggest that in many systems the effects of mutualistic interactions 
between ants and honeydew-producing Hemipterans, either positive or negative, are often 
assumed without ever being quantified.  A mutualism between the invasive ant Pheidole 
megacephala and the gray pineapple mealybug (Dysmicoccus neobrevipes), for instance, 
is thought to severely damage pineapple trees in Hawaii because the mealybug vectors a 
devastating phytopathogen (Beardsley et al. 1982).  Whether the interaction per se 
actually reduces plant fitness (i.e., whether ant-tending exacerbates the negative effects of 
the mealybugs), however, has not been tested.  Similarly, Way et al. (1999) surmise that 
the protection of pine trees by L. humile is partly a consequence of ant attendance of 
aphids, but they never specifically test this hypothesis.  In observational studies such as 
these (of which there are many), the assumed consequences of the ant-Hemipteran 
mutualisms may very well be real, but the lack of empirical data precludes any definitive 
conclusions or generalizations. 
I also argue that the ecological effects of ants and their interactions with other 
arthropods are simply ignored.  Ants may be noted as abundant predators, particularly in 
annual agricultural systems, but the consequences of their interactions with other 
 
55 
arthropods are often not quantified.  Several ant species, for example, tend the recently 
introduced aphid Aphis glycines in soybean (Glycine max) in North America (Ragsdale et 
al. 2004), but no effort has been made to determine the consequences of these mutualisms 
to the abundance and distribution of other natural enemies and herbivores, or to soybean 
yield.  Ecological consequences of ant-Hemipteran mutualisms are also often dismissed 
because they are assumed to be too local or too ephemeral to affect community structure 
or plant fitness.  Wimp and Whitham (2001) showed, however, that the local affects of an 
ant-aphid mutualism on the abundance and distribution of arthropods on individual trees 
also influenced arthropod abundance and diversity throughout the surrounding plant 
community.  Further, populations of some honeydew-producing Hemipteran taxa (e.g., 
aphids) may indeed be ephemeral, but even short-term ant-Hemipteran associations can 
have dramatic community-level consequences as demonstrated by Floate and Whitham 
(1994), and Kaplan and Eubanks (2005) (see above).  Ephemeral ant-Hemipteran 
associations may even induce long-term changes in host plant quality that affect other 
herbivores and their interactions with natural enemies later in the season (cf Van Zandt & 
Agrawal 2004). 
 
7. Conclusions 
Whether because they are assumed, ignored, dismissed, or simply overlooked, the 
ecological consequences of ant-Hemipteran mutualisms and their importance in 
arthropod communities remain poorly understood.  Mutualisms between ants and 
honeydew-producing Hemipterans are abundant and widespread in arthropod food webs 
and have potentially broad ecological effects because the presence of honeydew-
 
56 
producing Hemipterans dramatically alters the abundance and predatory behavior of ants 
on plants.  Future studies should focus on the ecological factors that influence the 
consequences of ant-Hemipteran mutualisms in order to provide greater insight into the 
role of positive species interactions in food web dynamics and greater predictability of 
the direct and indirect effects of herbivores and natural enemies on host plants.  Such 
predictability could substantially benefit biological control in agricultural systems with 
ants and Hemipterans.  Finally, I encourage a more concerted effort to study the 
consequences to arthropod communities and host plants of ant-Hemipteran mutualisms 
involving invasive ants, and to make comparisons with native ants.  Invasive ants have 
the potential to wreak ecological havoc in invaded habitats; thus, understanding the 
consequences of their interactions with Hemipterans could help predict and even manage 
their effects. 
 
References 
Banks, C. J. & Macaulay, E. D. M. 1967 Effects of Aphis fabae Scop. and of its attendant 
ants and insect predators on yields of field beans (Vicia faba L.). Ann. Appl. Biol. 
60, 445-453. 
Beardsley, J.W., Jr., Su, T.H., McEwen, F.L., & Gerling, D. 1982 Field investigations of 
the interrelationships of the big-headed ant, Pheidole megacephala, the gray 
pineapple mealybug, Dysmicoccus neobrevipes, and the pineapple mealybug wilt 
disease in Hawai?i, USA. Proc. Hawaiian Entomol. Soc. 24, 51-58. 
Beattie, A.J. 1985 The evolutionary ecology of ant-plant mutualisms. Cambridge: 
Cambridge University Press. 
 
57 
Bl?thgen, N., Verhaagh, M., Goit?a, W., Jaff?, K., Morawetz, W., & Barthlott, W. 2000 
How plants shape the ant community in the Amazonian rainforest canopy: the key 
role of extrafloral nectaries and homopteran honeydew. Oecologia 125, 229-240. 
Boucher, D.H., James, S. & Keeler, K.H. 1982 The ecology of mutualism. Annu. Rev. 
Ecol. Syst. 13, 315-347. 
Bradley, G.A. 1973 Effect of Formica obscuripes (Hymenoptera: Formicidae) on the 
predator-prey relationship between Hyperaspis congressis (Coleoptera: 
Coccinellidae) and Toumeyella numismaticum (Homoptera: Coccidae). Can. 
Entomol. 105, 1113-1118. 
Bronstein, J.L. 1994a Our current understanding of mutualism. Q. Rev. Biol. 69, 31-51. 
Bronstein, J.L. 1994b Conditional outcomes in mutualistic interactions. Trends Ecol. 
Evol. 9, 214-217. 
Buckley, R. 1983 Interaction between ants and membracid bugs decreases growth and 
seed set of a host plant bearing extrafloral nectaries. Oecologia 58, 132-136. 
Buckley, R.C. 1987 Interactions involving plants, Homoptera, and ants. Annu. Rev. Ecol. 
Syst. 18, 111-135. 
Buckley, R.C. & Gullan, P.J. 1991 More aggressive ant species (Hymenoptera: 
Formicidae) provide better protection for soft scales and mealybugs (Homoptera: 
Coccidae, Pseudococcidae). Biotropica 23, 282-286. 
Carroll, C.R. & Janzen, D.H. 1973 Ecology of foraging by ants. Annu. Rev. Ecol. Syst. 4, 
231-257. 
Carter, W. 1962 Insects in relation to plant disease. New York: John Wiley & Sons. 
 
58 
Cooper, L.B. 2005 The potential effects of red imported fire ants (Solenopsis invicta) on 
arthropod abundance and Cucumber Mosaic Virus. MSc Thesis, Auburn 
University, Auburn, AL, USA. 
Cushman, J.H. 1991 Host-plant mediation of insect mutualisms: variable outcomes in 
herbivore-ant interactions. Oikos 61, 138-144. 
Cushman, J.H. & Whitham, T.G. 1989 Conditional mutualism in a membracid-ant 
association: temporal, age-specific, and density-dependent effects. Ecology 70, 
1040-1047. 
Cushman, J.H. & Whitham, T.G. 1991 Competition mediating the outcome of a 
mutualism ? protective services of ants as a limiting resource for membracids. 
Am. Nat. 138, 851-865. 
Crutsinger, G.M. & Sanders, N.J. 2005 Aphid-tending ants affect secondary users in leaf 
shelters and rates of herbivory on Salix hookeriana in a coastal dune habitat. Am. 
Midl. Nat. 154, 296-304. 
Davidson, D.W. 1997 The role of resource imbalances in the evolutionary ecology of 
tropical arboreal ants. Biol. J. Linn. Soc. 61, 153-181. 
Davidson, D.W., Cook, S.C., Snelling, R.R., & Chua, T.H. 2003 Explaining the 
abundance of ants in lowland tropical rainforest canopies. Science 300, 969-972. 
Dejean, A., Bourgoin, T., & Gibernau, M. 1997 Ant species that protect figs against other 
ants: result of territoriality induced by a mutualistic homopteran. Ecoscience 4, 
446-453. 
Delabie, J.H.C. 2001 Trophobiosis between Formicidae and Hemiptera (Sternorrhyncha 
and Auchenorrhyncha): an overview. Neotrop. Entomol. 30, 501-516. 
 
59 
Del-Claro, K. & Oliveira, P.S. 1999 Ant-Homoptera interactions in a neotropical 
savanna: the honeydew-producing treehopper Guayaquila xiphias (Membracidae) 
and its associated ant fauna on Didymopanax vinosum (Araliaceae). Biotropica 
31, 135-144. 
Del-Claro, K. & Oliveira, P.S. 2000. Conditional outcomes in a neotropical treehopper-
ant association: temporal and species-specific variation in ant protection and 
homopteran fecundity. Oecologia 124, 156-165. 
Eubanks, M.D. 2001. Estimates of direct and indirect effects of red imported fire ants on 
biological control in field crops. Biol. Control 21,35-43. 
Eubanks, M.D., Blackwell, S.A., Parrish, C.J., Delamar, Z.D., & Hull-Sanders, H. 2002. 
Intraguild predation of beneficial arthropods by red imported fire ants in cotton. 
Environ. Entomol. 31,1168-1174. 
Floate, K. D. & Whitham, T.G. 1994 Aphid-ant interaction reduces chrysomelid 
herbivory in a cottonwood hybrid zone. Oecologia 97, 215-221. 
Fowler, S.V. & MacGarvin, M. 1985 The impact of hairy wood ants, Formica lugubris, 
on the guild structure of herbivorous insects on birch, Betula pubescens. J. Anim. 
Ecol. 54, 847-855. 
Fritz, R.S. 1983 Ant protection of a host plant?s defoliator: consequence of an ant-
membracid mutualism. Ecology 64, 789-797. 
H?lldobler, B. & Wilson, E.O. 1990 The ants. Cambridge, MA: Belknap Press. 
Holway, D.A., Lach, L., Suarez, A.V., Tsutsui, N.D., & Case, T.J. 2002 The causes and 
consequences of ant invasions. Annu. Rev. Ecol. Syst. 33, 181-233. 
 
60 
Ito, F., & Higashi, S. 1991 An indirect mutualism between oaks and wood ants via 
aphids. J. Anim. Ecol. 60, 463-470. 
Jutsum, A.R., Cherrett, J.M., & Fisher, M. 1981 Interactions between the fauna of citrus 
trees in Trinidad and the ants Atta cephalotes and Azteca sp. J. Appl. Ecol. 18, 
187-195. 
Kaneko, S. 2003 Different impacts of two species of aphid-tending ants with different 
aggressiveness on the number of emerging adults of the aphid?s primary 
parasitoid and hyperparasitoids. Ecol. Res. 18, 199-212. 
Kaplan, I., & Eubanks, M.D. 2002 Disruption of cotton aphid (Homoptera: Aphididae) ? 
natural enemy dynamics by red imported fire ants (Hymenoptera: Formicidae). 
Environ. Entomol. 31, 1175-1183. 
Kaplan, I., & Eubanks, M.D. 2005 Aphids alter the community-wide impact of fire ants. 
Ecology 86, 1640-1649. 
Karhu, K.J. 1998 Effects of ant exclusion during outbreaks of a defoliator and a sap-
sucker on birch. Ecol. Entomol. 23, 185-194. 
Kenne, M., Dji?to-Loridon, C., Orivel, J., Mony, R., Fabre, A. & Dejean, A. 2003 
Influence of insecticide treatments on ant-Hemiptera associations in tropical 
plantations. J. Econ. Entomol. 96, 251-258. 
Khoo, K. C. & Ho, C.T. 1992 The influence of Dolichoderus thoracicus (Hymenoptera: 
Formicidae) on losses due to Helopeltis theivora (Heteroptera: Miridae), black 
pod disease, and mammalian pests in cocoa in Malaysia. Bull. Entomol. Res. 82, 
485-491. 
 
61 
Lach, L. 2003 Invasive ants: unwanted partners in ant-plant interactions? Ann. Mo. Bot. 
Gard. 90, 91-108. 
L?hr, B. 1992 The pugnacious ant, Anoplolepis custodiens (Hymenoptera: Formicidae), 
and its beneficial effect on coconut production in Tanzania. Bull. Entomol. Res. 
82, 213-218. 
Messina, F.J. 1981 Plant protection as a consequence of an ant-membracid mutualism: 
interactions on goldenrod (Solidago sp.). Ecology 62, 1433-1440. 
Moya-Raygoza, G. & Nault, L.R. 2000 Obligatory mutualism between Dalbulus 
quinquenotatus (Homoptera: Cicadellidae) and attendant ants. Ann. Entomol. Soc. 
Am. 93, 929-940. 
Ness, J.H. & Bronstein, J.L. 2004 The effects of invasive ants on prospective ant 
mutualists. Biol. Invasions 6, 445-461. 
Nickerson, J.C., Rolph Kay, C.A., Buschman, L.L., & Whitcomb, W.H. 1977 The 
presence of Spissistilus festinus as a factor affecting egg predation by ants in 
soybeans. Fla. Entomol. 60, 193-199. 
Novgorodova, T.A. 2005 Ant-aphid interactions in multispecies ant communities: some 
ecological and ethological aspects. Eur. J. Entomol. 102, 495-501. 
Oliveira, P. S., & Del-Claro, K. 2005 Multitrophic interactions in a neotropical savanna: 
ant-hemipteran systems, associated insect herbivores and a host plant.  In Biotic 
interactions in the Tropics: their role in the maintenance of species diversity (eds. 
D. Burslem, M. Pinard, & S. Hartley), pp. 414-438.  New York: Cambridge 
University Press. 
 
62 
Ragsdale, D. W., Voegtlin, D. J., & O?Neil, R. J. 2004 Soybean aphid biology in North 
America. Ann. Entomol. Soc. Am. 97, 204-208. 
Rapp, G. & Salum, M.S. 1995 Ant fauna, pest damage and yield in relation to the density 
of weeds in coconut sites in Zanzibar, Tanzania. J. Appl. Entomol. 119, 45-48. 
Renault, C.K., Buffa, L.M. &  Delfino, M.A. 2005 An aphid-ant interaction: effects on 
different trophic levels. Ecol. Res. 20, 71-74. 
Rico-Gray, V. & Castro, G. 1996 Effect of an ant-aphid interaction on the reproductive 
fitness of Paullinia fuscecens (Sapindaceae). Southwest. Nat. 41, 434-440. 
Rico-Gray, V. & Thien, L.B. 1989 Ant-mealybug interaction decreases reproductive 
fitness of Schomburgkia tibicinis (Orchidaceae) in Mexico. J. Trop. Biol. 5, 109-
112. 
Rodriguez-Saona, C., Chalmers, J.A., Raj, S., & Thaler, J.S. 2005 Induced plant 
responses to multiple damagers: differential effects of an herbivore and its 
parasitoid. Oecologia 143, 566-577. 
Room, P.M. 1972 The fauna of the mistletoe Tapananthus bangwensis (Engl. & K. 
Krause) growing on cocoa in Ghana: relationships between fauna and mistletoe. J. 
Anim. Ecol. 41, 611-621. 
Rosenheim, J. A., L. R. Wilhoit, P. B. Goodell, E. E. Grafton-Cardwell, and T. F. Leigh. 
1997 Plant compensation, natural biological control, and herbivory by Aphis 
gossypii on pre-reproductive cotton: the anatomy of a non-pest. Entomol. Exp. 
Appl. 85, 45-63. 
Sipura, M. 1999 Tritrophic interactions: willows, herbivorous insects and insectivorous 
birds. Oecologia 121, 537-545. 
 
63 
Sipura, M. 2002 Contrasting effects of ants on the herbivory and growth of two willow 
species. Ecology 83, 2680-2690. 
Skinner, G.J. & Whittaker, J.B. 1981 An experimental investigation of inter-relationships 
between the wood-ant (Formica rufa) and some tree-canopy herbivores. J. Anim. 
Ecol. 50, 313-326. 
Stachowicz, J.J. 2001 Mutualism, facilitation, and the structure of ecological 
communities. BioScience 51, 235-246. 
Stechmann, D.-H., V?lkl, W, & Star?, P. 1996 Ant-attendance as a critical factor in the 
biological control of the banana aphid Pentalonia nigronervosa Coq. (Hom. 
Aphididae) in Oceania. J. Appl. Entomol. 120, 119-123. 
Stout, M.J., Workman, K.V., Bostock, R.M., & Duffey, S.S. 1998 Specificity of induced 
resistance in the tomato, Lycopersicon esculentum. Oecologia 113, 74-81. 
Suzuki, N., Ogura, K., & Katayama, N. 2004 Efficiency of herbivore exclusion by ants 
attracted to aphids on the vetch Vicia angustifolia L. (Leguminosae). Ecol. Res. 
19, 275-282. 
Tedders, W.L., Reilly, C.C., Wood, B.W., Morrison, R.K., & Lofgren C.S. 1990 
Behavior of Solenopsis invicta (Hymenoptera: Formicidae) in pecan orchards. 
Environ. Entomol. 19, 44-53. 
Van Mele, P. & Cuc, N.T.T. 2001 Farmers? perceptions and practices in use of 
Dolichoderus thoracicus (Smith) (Hymenoptera: Formicidae) for biological 
control of pests of sapodilla. Biol. Control 20, 23-29. 
Van Zandt, P.A. & Agrawal. A.A. 2004. Community-wide impacts of herbivore-induced 
plant responses in common milkweed Asclepias syriaca. Ecology 85, 2616-2629. 
 
64 
Vrieling, K., Smit, W., & van der Meijden, E. 1991 Tritrophic interactions between 
aphids (Aphis jacobaeae Schrank), ant species, Tyria jacobaeae L., and Senecio 
jacobaea L. lead to maintenance of genetic variation in pyrrolizidine alkaloid 
concentration. Oecologia 86, 177-182. 
Way, M.J. 1963 Mutualism between ants and honeydew-producing Homoptera. Annu. 
Rev. Entomol. 8, 307-344. 
Way, M.J. & Khoo, K.C. 1992 Role of ants in pest management. Annu. Rev. Entomol. 37, 
479-503. 
Way, M.J., Paiva, M.R. & Cammell, M.E. 1999 Natural biological control of the pine 
processionary moth Thaumetopoea pityocampa (Den & Schiff) by the Argentine 
ant Linepithema humile (Mayr) in Portugal. Agric. Forest Entomol. 1, 27-31. 
Wimp, G.M. & Whitham T.G. 2001 Biodiversity consequences of predation and host 
plant hybridization on an aphid-ant mutualism. Ecology 82, 440-452. 
Whittaker, J.B. & Warrington, S. 1985 An experimental field study of different levels of 
insect herbivory induced by Formica rufa predation on sycamore (Acer 
pseudoplatanus). III. Effects on tree growth. J. Appl. Ecol. 22, 797-811.
  
65 
T
a
bl
e 1.
  S
u
m
m
a
r
y
 o
f
 25 s
t
ud
i
e
s
 t
h
at
 ex
p
l
i
c
i
t
l
y
 t
e
s
t
 f
o
r
 ef
f
ect
s
 o
n
 pl
a
n
t
s
 
o
f
 
m
u
t
u
a
l
i
s
t
i
c 
i
n
t
e
r
act
i
o
n
s
 
b
e
t
w
een
 a
n
t
s
 an
d 
h
o
n
e
y
d
ew-
p
r
o
duc
i
n
g He
m
i
p
t
e
r
a
ns
.
  Negat
i
ve e
f
f
e
c
t
s
 i
n
d
i
cat
e i
n
cr
eas
ed p
l
a
n
t
 dam
a
ge an
d/
o
r
 decr
eas
ed pl
a
n
t
 gr
o
w
t
h
 an
d 
r
e
pr
o
duct
i
o
n
.
  P
o
s
i
t
i
v
e
 e
f
f
ect
s
 
i
n
d
i
cat
e decr
eas
e
d pl
a
n
t
 da
m
a
ge a
n
d/o
r
 i
n
cr
eas
ed p
l
a
n
t
 gr
o
w
t
h
 an
d r
e
pr
o
duc
t
i
o
n
.
   
A
n
t S
p
e
c
i
e
s
 (
s
ta
tu
s
a
)
 
H
e
m
i
p
t
e
r
a
n
 S
p
eci
es
 
S
y
s
t
em
 
H
o
s
t
 p
l
an
t
 
M
e
t
h
o
d
b
 P
l
a
n
t
 
t
r
ai
t 
me
as
u
r
e
d
 
C
i
t
a
tio
n
 
N
e
g
a
t
iv
e
 ef
f
ect
 
  
 
 
 
 
L
a
s
i
us
 ni
ge
r
 (N
) 
A
phi
s
 f
a
b
a
e
 
M
a
n
a
ge
d:
 f
i
e
l
d c
r
o
p
 
F
a
b
a
 b
e
a
n
 
V
i
c
i
a f
a
ba
 
E
xp(1) 
# v
i
a
b
l
e
 s
e
e
d
s
 
1 
I
r
id
o
m
yr
m
e
x 
sp
. (
N
)
 S
e
x
t
i
u
s
 
v
i
re
sc
e
n
s 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
A
c
ac
ia 
A
c
ac
i
a
 de
c
u
r
r
e
n
s
 
E
x
p
(
3
)
 
P
l
a
n
t
 gr
ow
t
h
,
 # 
s
eed
s
 
2 
Cr
e
m
at
og
as
t
e
r
 br
e
v
i
s
pi
nos
a
 
(N
) 
P
l
anoc
o
c
c
u
s
 c
i
t
r
i
 
N
a
tu
r
a
l
:
 c
o
as
tal
 d
u
n
e
 
O
r
c
h
id
 
Sc
hom
bur
g
k
i
a
 t
i
bi
c
i
n
i
s
 
E
xp(1) 
P
l
a
n
t
 d
a
m
a
ge
,
  
# f
l
ow
e
r
s
,
 #
 f
r
u
i
t
s
 
3 
Cam
ponot
us
 p
l
anat
us
 (
N
)
 
U
n
id
e
n
tif
ie
d
 
ap
h
i
d
 s
p
. 
N
a
tu
r
a
l
:
 c
o
as
tal
 d
u
n
e
 
L
i
a
n
a 
P
a
u
llin
ia
 fu
s
c
ec
en
s
 
E
xp(1) 
# f
l
ow
e
r
s
 w
i
t
h
 s
e
e
d
s
,
 
#
 s
eed
s
 p
e
r
 f
l
o
w
er
 
4 
F
o
r
m
i
c
a aq
ui
l
oni
a
 (N
) 
My
r
m
i
c
a r
ubr
a
 (N
) 
P
t
er
o
c
o
m
m
a
 s
a
l
i
cis
 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
W
illo
w
 
Sal
i
x
 m
y
r
s
i
ni
f
o
l
i
a 
E
xp(3) 
L
e
a
f
 da
m
a
ge
,
  
p
l
a
n
t
 gr
ow
t
h
 
5 
Cam
ponot
us
 s
p
.
 (N
) 
A
phi
s
 c
o
r
e
ops
i
d
i
s
 
Na
t
u
r
a
l
:
 g
r
a
ssl
a
n
d
 
Ast
e
r
 
B
i
de
n
s
 pi
l
o
s
a
 
E
xp(1) # 
s
e
e
d
s
 
6 
  
66 
Sol
e
n
ops
i
s
 i
n
v
i
c
t
a
 (I) 
U
n
i
d
e
n
t
i
f
i
e
d 
a
p
h
i
d
 
s
pp.
 
M
a
n
a
ge
d:
 v
e
ge
t
a
b
l
e
 
cr
o
p
 
To
mato
 
Lycope
r
s
i
c
on
 e
s
cu
l
e
nt
um
 
E
xp(1) 
L
e
v
e
l
 of
 i
n
f
e
c
t
i
o
n
  
by
 pl
a
n
t
 v
i
r
u
s
 
7 
N
o
 e
ffe
c
t
 
  
 
 
 
 
F
o
r
m
i
c
a s
ubs
e
r
i
c
e
a
 (N
) 
V
anduz
c
a
 ar
q
uat
a 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
B
l
ac
k
 lo
c
u
s
t
 
R
obi
n
i
a
 ps
e
u
d
oac
a
c
i
a
 
Co
rr
 
L
e
a
f
 da
m
a
ge
,
  
p
l
a
n
t
 gr
ow
t
h
 
8 
B
r
ac
h
y
m
y
r
m
e
x
 obs
c
u
r
i
or
 (N
) 
D
a
l
bul
us
 
qui
nqu
e
n
o
t
at
us
 
Na
t
u
r
a
l
:
 g
r
a
ssl
a
n
d
 
Ga
m
a
g
r
a
ss 
Tw
o
 
Tr
i
p
s
a
c
u
m
 sp
ec
i
e
s 
E
xp(1) 
# l
e
a
v
e
s
,
 pl
a
n
t
 
gr
o
w
t
h
 
9 
P
o
s
i
ti
v
e
 e
ffe
c
t
 
 
  
 
 
 
Cr
e
m
at
og
as
t
e
r
 af
r
i
c
ana
 (N
) 
C
a
t
a
e
noc
occ
u
s
 l
o
r
a
n
t
h
i
 
St
i
c
t
o
co
ccu
s sj
ost
e
d
t
i
 
M
a
n
a
ge
d:
 c
o
c
o
a
 
p
l
a
n
ta
tio
n
 
M
i
s
tle
to
e
 
Tapi
n
a
n
t
hus
 b
angw
e
n
s
i
s
 
E
xp(1) L
e
a
f
 
da
m
a
ge
,
 
p
l
a
n
t
 gr
ow
t
h
 
10 
Az
te
ca
 s
p
.
 (N
) 
C
o
c
c
u
s
 v
i
ri
d
i
s 
M
a
n
a
ge
d:
 o
r
c
h
a
r
d 
O
r
a
n
ge
 
C
itr
u
s
 r
e
t
i
cu
la
ta
 
E
xp(1) L
e
a
f
 
da
m
a
ge
 
11 
Fo
r
m
i
c
a
 s
p
.
 (N
) 
P
ubl
i
l
i
a
 c
o
n
c
av
a 
N
a
t
u
ra
l
:
 o
l
d f
i
e
l
d 
G
o
l
d
e
n
r
o
d 
Sol
i
d
ago al
t
i
s
s
i
m
a
 
C
o
r
r
 
P
l
a
n
t
 gr
ow
t
h
,
 # 
s
eed
s
 
12 
F
o
rmi
c
a
 ru
f
a
 (N
) 
P
e
r
i
phy
l
l
u
s
 
tes
t
u
d
in
a
ceu
s
 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
S
y
c
a
mo
r
e
 
A
c
e
r
 ps
e
udop
l
a
t
a
n
u
s
 
E
xp(1) L
e
a
f
 
da
m
a
ge
 
13 
F
o
rmi
c
a
 ru
f
a
 (N
) 
P
e
r
i
phy
l
l
u
s
 
tes
t
u
d
in
a
ceu
s
 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
S
y
c
a
mo
r
e
 
A
c
e
r
 ps
e
udop
l
a
t
a
n
u
s
 
E
x
p
(
1
)
 
P
l
a
n
t
 gr
ow
t
h
,
 # 
s
eed
s
 
14 
  
67 
F
o
rmi
c
a
 y
e
sse
n
s
i
s
 (N
) 
T
ube
r
c
u
l
at
us
 
que
r
c
i
c
o
l
a 
N
a
t
u
ra
l
:
 c
h
a
p
a
rra
l
 
O
a
k 
Q
u
e
r
c
u
s
 de
n
t
at
a
 
Co
rr
 # 
a
c
o
r
n
s
,
 
a
c
o
r
n d
a
m
a
ge
 
15 
D
o
l
i
c
hode
r
u
s
 t
h
or
ac
i
c
us
 (N
) 
Cat
a
e
noc
oc
c
u
s
 
hi
s
p
i
dus
 
M
a
n
a
g
e
d
:
 p
l
a
n
ta
tio
n
 
C
o
co
a 
T
h
e
obr
om
a c
a
c
a
o
 
E
xp(1) F
r
u
i
t
 
da
m
a
ge
 
16 
A
nopl
ol
e
p
i
s
 c
u
s
t
odi
e
n
s
 (N
) 
C
e
r
a
ta
p
h
is
 va
r
i
a
b
il
is
 
M
a
n
a
g
e
d
:
 p
l
a
n
ta
tio
n
 
C
o
co
n
u
t 
Coc
o
s
 nuc
i
f
e
r
a 
C
o
r
r
 
P
a
lm
 
le
s
i
o
n
s
,
 #
 n
u
ts
,
 
 
n
u
t
 da
m
a
ge
 
17 
Fo
r
m
i
c
a
 s
p
.
 (N
) 
Chai
t
ophor
us
 
popul
i
c
o
l
a 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
C
o
tto
n
w
o
o
d
 
P
opul
us
 f
r
e
m
ont
i
i
 
E
xp(2) L
e
a
f
 
da
m
a
ge
 
18 
Cam
ponot
us
 br
ut
us
 (N
) 
H
i
l
d
a
 unda
t
a
 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
F
i
g
 
F
i
c
u
s
 v
a
l
l
i
s
c
h
ouda
e
 
Co
rr
 F
r
u
i
t
 
da
m
a
ge
 
19 
F
o
r
m
i
c
a aq
ui
l
oni
a
 (N
) 
Sy
m
y
do
bi
us
 obl
ongus
 N
a
tu
r
a
l
:
 
w
o
o
d
la
n
d
 
B
i
r
c
h
 
B
e
t
u
l
a
 pub
e
s
c
e
ns
 
E
xp(1) L
e
a
f
 
da
m
a
ge
 
20 
D
o
l
i
c
hode
r
u
s
 t
h
or
ac
i
c
us
 (N
) 
P
l
anoc
o
c
c
u
s
 l
i
l
a
c
i
nus
 
M
a
n
a
g
e
d
:
 p
l
a
n
ta
tio
n
 
S
a
p
o
d
ill
a 
Mani
l
k
ar
a
 z
apot
a 
E
xp(1) F
r
u
i
t
 
da
m
a
ge
 
21 
F
o
r
m
i
c
a aq
ui
l
oni
a
 (N
) 
My
r
m
i
c
a r
ubr
a
 (N
) 
P
t
er
o
c
o
m
m
a
 s
a
l
i
cis
 
N
a
tu
r
a
l
:
 w
o
o
d
la
n
d
 
W
illo
w
 
Sal
i
x
 p
h
y
l
i
c
i
f
ol
i
a
 
E
xp(3) 
L
e
a
f
 da
m
a
ge
,
  
p
l
a
n
t
 gr
ow
t
h
 
5 
L
a
s
i
us
 j
a
poni
c
u
s
 (N
) 
T
e
t
r
am
or
i
u
m
 t
s
us
hi
m
a
e
 (N
) 
C
o
mp
le
x
 o
f
 th
r
e
e
 
a
p
h
i
d
 s
p
eci
es
d
 
Na
t
u
r
a
l
:
 g
r
a
ssl
a
n
d
 
Ve
t
c
h
 
V
i
c
i
a
 angus
t
i
f
ol
i
a
 
Co
rr
 # 
s
e
e
d
s
 
22 
  
68 
C
o
m
p
l
e
x
 o
f
 f
i
v
e
 s
p
eci
es
c
 (N
) 
G
uay
aqu
i
l
a
 x
i
ph
i
a
s
 
N
a
t
u
ra
l
:
 c
e
rra
do
 
L
i
a
n
a
 
D
i
dy
m
opa
nax
 v
i
nos
um
 
E
xp(1) 
D
a
m
a
ge
 t
o
 l
e
a
v
e
s
 
a
n
d
 ap
ic
al
 
me
r
i
s
t
e
m
s
 
23 
F
o
rmi
c
a
 o
b
s
c
u
ri
p
e
s 
(N
)
 
U
n
id
e
n
tif
ie
d
 
A
phi
s
 
sp
.
 
N
a
tu
r
a
l
:
 c
o
as
tal
 d
u
n
e
 
W
illo
w
 
Sal
i
x
 h
ook
e
r
i
a
n
a
 
Co
rr
 L
e
a
f
 
da
m
a
ge
 
24 
Sol
e
n
ops
i
s
 i
n
v
i
c
t
a
 (I) 
A
phi
s
 gos
s
y
p
i
i
 
M
a
n
a
ge
d:
 f
i
e
l
d c
r
o
p
 
Co
t
t
o
n
 
G
o
s
s
y
pi
um
 hi
r
s
u
t
u
m
 
E
xp(2) 
# c
o
t
t
o
n s
q
ua
r
e
s
,
 bo
l
l
s
,
 
seed
s;
 s
eedc
o
t
t
o
n m
a
ss 
25 
a
 S
t
a
t
u
s
 
o
f
 a
n
t
 a
s
 a
 n
a
t
i
v
e
 (N
) 
o
r
 i
n
v
a
si
v
e
 (I) s
p
e
c
i
e
s
 a
t
 s
t
u
d
y
 si
t
e
. 
b 
S
t
ud
i
e
s
 cat
ego
r
i
zed as
 e
i
t
h
er
 co
r
r
e
l
at
i
v
e,
 ?
C
o
r
r
?
,
 
or
 ex
per
i
m
e
n
t
a
l
,
 ?
E
x
p
?
.
  E
x
per
i
m
e
n
t
a
l
 s
t
udi
es
 ar
e 
f
u
r
t
h
e
r
 cl
a
s
s
i
f
i
e
d
 by
 t
h
e 
m
a
ni
pu
l
a
t
i
o
n
 e
m
p
l
o
y
ed:
 (
1
)
 an
t
 ex
c
l
u
s
i
o
n
 a
l
o
n
e,
 (
2
)
 Hem
i
p
t
e
r
a
n
 e
x
c
l
us
i
o
n
 a
l
o
n
e,
 (
3
)
 b
o
t
h
 an
t
 an
d Hem
i
pt
er
an
 exc
l
u
s
i
o
n
 
i
n
 a 
c
r
os
s
e
d
 d
e
s
i
gn
.
  T
h
e
 s
i
gn
i
f
i
can
c
e
 of
 th
i
s
 di
f
f
e
r
en
c
e
 i
n
 
m
e
th
od
ol
ogy
 
i
s
 di
s
c
u
s
s
e
d
 
i
n
 th
e
 te
x
t
.
 
c
 C
a
mponotus
 cr
a
s
s
u
s
,
 C
.
 r
e
ngger
i
,
 C
.
 r
u
f
i
pes
,
 C
e
phalotes
 pus
illus
,
 
an
d 
E
c
tatomma e
d
e
n
tatum 
d
 
A
phi
s
 cr
acci
vor
a
, 
A
cyr
thos
iphon pis
u
m
, a
n
d
 
M
e
gour
a cr
as
s
i
cauda
. 
C
i
t
a
t
i
o
n
s
:
  1-
B
a
n
k
s
 &
 M
acau
l
a
y
 1976;
 2-
B
u
ck
l
e
y
 
1983;
 3-
R
i
co
-
G
r
a
y
 &
 T
h
i
e
n
 1989
;
 4-
R
i
co
-
G
r
a
y
 &
 
C
a
s
t
r
o
 1996;
 5-
S
i
pur
a 
2002;
 6-
R
e
n
a
u
l
t
 et
 al
.
 2005;
 7-
C
o
o
p
er
 2005;
 8-
F
r
i
t
z 1983;
 9-
M
o
y
a
-
R
a
y
go
za &
 Nau
l
t
 2000;
 10-
R
o
o
m
 1972;
 11-
J
u
t
s
um
 et
 al
.
 
1981;
 12-
M
e
s
s
i
na 1981;
 13-
Sk
i
n
n
e
r
 &
 
Whi
t
t
a
ker
 1
981;
 14-
Whi
t
t
a
ker
 &
 W
a
r
r
i
n
gt
o
n
 1985;
 15-
I
t
o
 &
 H
i
ga
s
h
i
 1991
;
 16-
Kh
o
o
 &
 
  
69 
Ho
 1992;
 17-
L
?
h
r
 1992;
 18-
Fl
o
a
t
e
 &
 W
h
i
t
h
a
m
 19
94;
 19-
De
j
e
a
n
 et
 al
.
 1997;
 20-
K
a
r
h
u 1998;
 21-
Van M
e
l
e
 
&
 C
u
c 2001;
 22-
Suzuk
i
 et
 al
.
 2004;
 23-
O
l
i
v
e
i
r
a
 &
 De
l
-
C
l
ar
o
 2005;
 
24-
C
r
ut
s
i
n
g
er
 &
 Sa
n
d
er
s
 2005;
 25-
C
h
apt
e
r
 3.
  
 
70 
Figure 1.  Potential interactions involving a mutualism between ants and honeydew-
producing Hemipterans.  Arrows indicate the direction of effects, positive (+) and 
negative (-).  Note that the net effect of the mutualism on the host plant may be positive 
or negative depending on the ratio between the direct cost to the host plant of Hemipteran 
herbivory and the indirect benefit to the host plant of ant suppression of herbivory by 
other herbivores. 
 
 
 
 
 
 
71 
Figure 2.  Conditionality in the consequences to a host plant of an ant-Hemipteran 
mutualism based on the intensity of non-Hemipteran herbivory.  The cost to the plant of 
the mutualism (in terms of loss of fitness caused by plant damage by Hemipterans) is 
shown as the dotted curve and increases with increasing Hemipteran density.  The benefit 
to the plant of the mutualism (in terms of increased plant fitness caused by ant 
suppression of plant damage by other herbivores) is shown as the two solid curves.  At a 
high level of non-Hemipteran herbivory, the benefit of the mutualism to plant fitness is 
predicted to outweigh the cost at all but extremely high levels of Hemipteran density.  At 
a low level of non-Hemipteran herbivory, however, the mutualism either provides only a 
slight fitness benefit (at lower Hemipteran densities) or results in a net loss of fitness (at 
higher Hemipteran densities). 
 
 
72 
CHAPTER THREE: 
A FACULTATIVE MUTUALISM BETWEEN COTTON APHIDS 
AND AN INVASIVE ANT INDIRECTLY BENEFITS 
PLANT REPRODUCTION 
 
Abstract.  Ants and aphids routinely engage in mutualistic interactions in which 
aphids barter a food resource (honeydew) for protection from predators by honeydew-
collecting ants.  The consequences of ant-aphid mutualisms to plants, particularly those 
involving invasive ant species, are poorly studied but could be negative or positive 
depending on the cost of herbivory by ant-tended aphids relative to the benefit of 
increased ant suppression of other herbivores.  I conducted field and greenhouse 
experiments over two years in which I manipulated the presence and absence of cotton 
aphids (Aphis gossypii) on cotton plants to test the hypothesis that a mutualism between 
cotton aphids and an invasive ant, the red imported fire ant (Solenopsis invicta) benefits 
cotton plants by increasing fire ant suppression of beet armyworm caterpillars 
(Spodoptera exigua).  I also manipulated caterpillar density to test whether the benefit of 
the mutualism varied with the level of caterpillar damage to plants.  Greater numbers of 
fire ants foraged on plants with cotton aphids than on plants without cotton aphids, 
resulting in a significant reduction in caterpillar survival and caterpillar herbivory of 
leaves, squares, and bolls on plants with aphids.  Consequently, cotton aphids indirectly 
 
73 
increased cotton reproduction.  In one field experiment, plants with cotton aphids 
produced on average 16% more bolls, 25% more seeds, and 10% greater seedcotton mass 
than plants without aphids.  The indirect benefit of cotton aphids, however, varied with 
caterpillar density:  the number of bolls per plant at harvest was 32% greater on plants 
with aphids than on plants without aphids at high caterpillar density versus just 3% 
greater on plants with aphids at low caterpillar density.  My results highlight the potential 
benefit to an economically important plant of a mutualism between an aphid pest and an 
invasive ant species.  Further, my results provide the first experimental evidence that the 
consequences to plants of ant-aphid mutualisms are context-dependent and can vary at 
different densities of non-aphid herbivores. 
 
INTRODUCTION 
Ants commonly engage in mutualistic interactions with honeydew-producing 
Hemipteran insects, including aphids, scales, mealybugs, and Membracid treehoppers 
(Way 1963, H?lldobler and Wilson 1990).  Ants are attracted to honeydew as a 
predictable, renewable food resource and subsequently tend honeydew-producing 
Hemipterans, protecting them from predators and parasitoids (Way 1963, Buckley 1987, 
H?lldobler and Wilson 1990).  From the perspective of the plants hosting these 
interactions, ant-Hemipteran mutualisms can be either harmful or beneficial.  Increased 
Hemipteran abundance in the presence of tending ants can result in decreased plant 
growth and reproduction because Hemipterans feed from plant phloem and often vector 
bacterial and viral pathogens (Buckley 1987, Delabie 2001).  Hemipteran-tending ants, 
however, exhibit increased aggressiveness towards other arthropods on the plant (Way 
 
74 
1963), and thereby potentially protect plants from other, more damaging herbivores (e.g., 
Nickerson 1977, Cushman et al. 1998).  The net consequence of an ant-Hemipteran 
mutualism to its host plant, therefore, could be negative or positive depending on the ratio 
of the direct cost of feeding by Hemipterans to the indirect benefit of increased ant 
suppression of other herbivores (Carroll and Janzen 1973, Fritz 1983, Buckley 1987, 
Lach 2003). 
Surprisingly few studies have investigated the consequences of ant-Hemipteran 
mutualisms to plants considering the ubiquity of these interactions in terrestrial 
communities.  Ant-Hemipteran mutualisms occur on most types of plants (including 
grasses, forbs, vines, shrubs, and trees) in most plant families from temperate to tropical 
latitudes (Buckley 1987, Bl?thgen et al. 2000, Delabie 2001).  Ant-Hemipteran 
mutualisms also commonly occur in managed (e.g., agricultural) as well as in natural 
systems (Buckley 1987, Way and Khoo 1992).  In a review of the relatively few studies 
that explicitly investigated the effects on plants of mutualistic interactions between ants 
and Hemipterans, Styrsky and Eubanks (in review) found that the majority of studies 
documented positive effects in terms of decreased plant damage by non-Hemipteran 
herbivores and/or increased plant growth and reproduction.  Messina (1981), for instance, 
reported that Membracid-tending Formica ants prevented defoliation of goldenrod plants 
by two Chrysomelid beetles, resulting in greater plant growth and seed production 
relative to neighboring plants that did not host Membracid aggregations.  Similarly, 
mealybug-tending Dolichoderus ants protected cocoa pods and sapodilla fruits from 
damage by Mirids (Heteroptera) (Khoo and Ho 1992) and Pyralid larvae (Lepidoptera) 
(Van Mele and Cuc 2001), respectively, without any direct yield loss caused by the 
 
75 
mealybugs.  Other studies, however, show that increased Hemipteran abundance in the 
presence of tending ants decreases plant fitness.  Renault et al. (2005), for example, 
reported that the aster Bidens pilosa produced significantly fewer viable seeds on plants 
with ant-tended aphids compared to plants from which ants had been excluded. 
All previous studies that specifically investigated the effects of ant-Hemipteran 
mutualisms on plant fitness have focused on mutualisms involving only native species of 
ants (Styrsky and Eubanks, in review).  Invasive ants, that subset of unintentionally 
introduced ant species that establish populations in natural areas undisturbed by human 
activity, also tend honeydew-producing Hemipterans and may disrupt mutualisms 
involving native ants (Holway et al. 2002, Lach 2003, Ness and Bronstein 2004).  
Because of their comparatively greater abundance and aggression in invaded habitats, 
invasive ants may have much stronger indirect effects on plants, either positive or 
negative, relative to native ants.  Invasive ants are potentially costly to plants because 
they are often associated with local Hemipteran outbreaks (Holway et al. 2002, Ness and 
Bronstein 2004), but they may also more effectively suppress other herbivores (Way et 
al. 1999, Lach 2003).  Studies investigating the effects on plants of ant-Hemipteran 
mutualisms involving invasive ants, therefore, are needed to better predict and manage 
the ecological consequences of invasive ants in both natural and managed systems. 
Predictability of the consequences to plants of ant-Hemipteran mutualisms 
involving either both native and invasive ants may also be improved by increased 
understanding of how the consequences vary in different ecological contexts (Lach 
2003).  Only one study has specifically investigated conditionality in the effects of ant-
Hemipteran mutualisms on plants.  Sipura (2002) showed that the net consequence to 
 
76 
plants of a mutualism between the ant Formica aquilonia and the aphid Pterocomma 
salicis differed between two species of willow based on differences in foliar 
concentrations of phenolic glycosides, a trait that confers resistance against leaf-chewing 
herbivores.  As a consequence of increased ant suppression of other herbivores, the ant-
aphid mutualism resulted in decreased leaf damage and increased growth in the willow 
species with low resistance.  In contrast, the ant-aphid mutualism negatively affected 
growth in the willow species with high resistance, presumably because the cost of aphid 
herbivory outweighed the very limited benefit of ant suppression of leaf-chewing insects. 
Our understanding of the consequences of ant-Hemipteran mutualisms to plants is 
further limited by the fact that almost all previous studies have focused on mutualisms 
hosted by woody species of plants, primarily trees and shrubs in natural woodlands and, 
secondarily, trees of agricultural importance in orchards and plantations (Eubanks and 
Styrsky, in press).  Ant-Hemipteran mutualisms may be more beneficial to host plants in 
woodlands and orchards because the relative stability of these habitats may promote long-
term establishment of both ant and Hemipteran colonies (Way and Khoo 1992).  Further, 
annual herbaceous plants may be particularly vulnerable to fitness costs associated with 
Hemipteran herbivory because of the allocation of resources to a single reproductive 
bout.  Whether ant-Hemipteran mutualisms indirectly harm or benefit more ephemeral, 
herbaceous plants or plants in more ephemeral agricultural systems such as grain, fiber, 
or vegetable crops is practically unknown. 
I present here the results of field experiments designed to investigate the 
consequences of a facultative mutualism between an invasive ant species, the red 
imported fire ant (Solenopsis invicta), and cotton aphids (Aphis gossypii) to an annual 
 
77 
host plant of agricultural importance, cotton (Gossypium hirsutum).  I asked two 
questions: (1) Can the fire ant-cotton aphid mutualism indirectly benefit cotton 
reproduction as a consequence of increased fire ant predation of a leaf-chewing 
herbivore, the beet armyworm caterpillar (Spodoptera exigua)?, and (2) Does the 
outcome of the fire ant - cotton aphid mutualism to cotton reproduction vary with 
caterpillar density?  I predicted that fire ant suppression of caterpillars would be greater 
on plants hosting cotton aphids than on plants without aphids, resulting in decreased 
caterpillar damage and increased production of cotton bolls (cotton fruit) on plants with 
aphids.  Because the indirect benefit of increased fire ant predation of caterpillars is more 
likely to outweigh the direct cost of cotton aphids when caterpillar density (hence, 
potential caterpillar damage to plants) is high, I further predicted that boll production 
would be greater on plants with aphids than on plants without aphids at high caterpillar 
density but not at low caterpillar density. 
 
METHODS 
Study organisms 
Red imported fire ants (?fire ants? from here forward) were unintentionally 
introduced into the United States through the port of Mobile, Alabama approximately 75 
years ago (Vinson 1997).  They have since spread throughout the Gulf Coast states, east 
to the coastal areas of South Carolina, North Carolina, and Virginia, and were secondarily 
introduced to California where they continue to expand their range northward (Vinson 
1997, Korzukhin et al. 2001).  Fire ants are notorious for their pervasive negative effects 
on invertebrate and vertebrate fauna in invaded habitats, effects typically attributed to 
 
78 
their broad omnivory, extreme aggression, and superabundance (Vinson 1997, Wojcik et 
al. 2001, Holway et al. 2002).  Fire ants are important predators in agricultural systems in 
particular where they can exert effective biological control of several herbivore pests, but 
also attack or disrupt other beneficial predators (Vinson 1997, Eubanks 2001, Eubanks et 
al. 2002). 
Fire ants forage primarily on the ground but will readily climb plants to tend 
honeydew-producing Hemipterans (Ness and Bronstein 2004), including cotton aphids on 
cotton plants (Kaplan and Eubanks 2002, Diaz et al. 2004).  Cotton aphids injure cotton 
plants by continuously feeding on plant phloem, which can result in stunted growth, loss 
of leaf surface area, premature leaf abscission, lower boll (fruit) retention, and reduced 
cotton lint mass (Fuchs and Minzenmayer 1995, Rosenheim et al. 1997).  Cotton aphids 
are effectively protected from other predators by tending fire ants (Kaplan and Eubanks 
2002); thus, the presence of fire ants on plants potentially exacerbates the negative effects 
of cotton aphids on cotton plant growth and reproduction.  Cotton plants are fairly 
tolerant to cotton aphid herbivory (Rosenheim et al. 1997), however, and could indirectly 
benefit from aphid-tending fire ants.  Fire ants aggressively patrol cotton plants in the 
presence of cotton aphids and significantly reduce the abundance of more damaging, non-
aphid herbivores including caterpillars and Lygus plant bugs (Diaz et al. 2004, Kaplan 
and Eubanks 2005).  The model non-aphid herbivore used in this study, the beet 
armyworm caterpillar, is considered a secondary pest of cotton, but in sufficient numbers 
it can threaten plant reproduction by defoliating plants and feeding on developing cotton 
squares (flower buds), flowers, and bolls (Mascarenhas et al. 1999).  Cotton aphids and 
 
79 
beet armyworm caterpillars can be found together on cotton plants throughout the 
growing season. 
 
Field experiments 
 I conducted the field experiments in 2003 and 2004 at the E. V. Smith Research 
Center of the Auburn University Agricultural Experiment Station in Macon and Elmore 
Counties, Alabama, USA.  The field experiment in each year consisted of a 2 x 2 
factorial, randomized block design in which I manipulated cotton aphid density (aphids 
present or excluded) and beet armyworm caterpillar density (low or high) on cotton 
plants in 1.8 x 1.8 x 1.8-m field cages.  The field cages were constructed of 32 x 32 mesh 
Lumite screening (opening size 530 ?m; Bioquip, Gardena, California, USA) that 
prevented aphid movement in or out of the cages.  I randomly assigned each treatment 
combination to one of four cages in each of seven blocks (n = 28 cages total) in an 8-ha 
field of conventional cotton in Macon County in 2003.  I followed the same procedure in 
2004 but with eight additional cages in two additional blocks (n = 36 cages total) in a 4-
ha field of conventional cotton in Elmore County.  Densities of fire ants were high along 
field margins in both years and additional colonies established naturally within the fields 
over the course of the growing seasons.  The blocks were arranged linearly in both years 
so that each cage was approximately 35 m from a field edge. 
I erected each field cage over two rows of cotton seedlings before they were 
colonized by any insects.  In 2003, I thinned the seedlings in each cage to ten evenly 
spaced plants per row.  By the middle of the season, however, branches on neighboring 
plants overlapped so extensively that sampling individual plants became very difficult, as 
 
80 
did moving around inside the cages without disturbing the plants (and fire ants).  In 2004, 
therefore, I removed all the seedlings in one of the two caged rows to provide space in 
which to work and thinned the seedlings in the remaining row to five evenly spaced 
plants.  The bottom edges of the cages were buried to block movement of all arthropods 
except fire ants.  Fire ants were able to access the caged plants via foraging tunnels they 
constructed under the cage walls. 
 I established the experimental treatments in late June 2003 and 2004, when cotton 
aphids appeared naturally in the fields.  I manipulated cotton aphid density by 
transferring to each plant in half of the cages approximately 300 aphids collected from 
the surrounding field (aphid-inclusion treatment).  In the absence of predators, aphid 
colonies were quickly established on caged plants.  No aphids were added to plants in the 
remaining cages (aphid-exclusion treatment).  Unfortunately, in 2003 cotton aphids 
colonized several of the ?aphid-exclusion? cages, and predatory larvae of ladybird beetles 
and lacewings infiltrated several of the ?aphid-inclusion? cages, causing aphid population 
crashes.  This resulted in a range of cotton aphid densities in cages rather than the strict 
presence or absence of aphids (see below).  Cotton aphids and aphid predators 
occasionally colonized cages in 2004, but by more frequently inspecting plants and 
removing unwanted individuals, I was able to maintain the two aphid density treatments. 
I manipulated caterpillar density by transferring to plants caterpillars reared in the 
lab from commercially purchased eggs (Benzon Research, Carlisle, Pennsylvania, USA).  
Cages in the low-density and high-density treatments received 5 and 15 caterpillars per 
plant per week, respectively, in 2003.  These densities are within the range of what is 
normally encountered in the field.  Beet armyworm caterpillars were reared on artificial 
 
81 
diet (Product #F9219B, Bio-Serv, Frenchtown, NJ) until they reached their third instar 
before transferring them to several haphazardly chosen leaves on the caged plants.  
Incidental caterpillar mortality was very high, however, because caterpillars apparently 
had difficulty acclimating to walking on leaves (they repeatedly fell from the plants) and 
feeding on cotton foliage.  Consequently, I could not effectively maintain two levels of 
caterpillar density in 2003.  In 2004, I transferred beet armyworm caterpillars to plants 
immediately after the caterpillars hatched at rates of 10 and 30 caterpillars per plant per 
week in the low-density and high-density treatments, respectively.  I used higher densities 
of caterpillars in 2004 to ensure that at least some caterpillars survived to begin feeding 
on the cotton foliage.  Again, these caterpillar densities are representative of what occur 
naturally; beet armyworm caterpillar eggs are typically oviposited in clusters of 
approximately 50 to 200 eggs (Ruberson et al. 1994).  The 10 or 30 caterpillars were 
transferred en masse to the underside of a single, haphazardly chosen, mainstem leaf in 
the middle of the plant canopy to mimic the oviposition behavior of the female beet 
armyworm moth (Ruberson et al. 1994).  This procedure protected the caterpillars from 
heavy dew accumulations on the upper surfaces of leaves and allowed them to acclimate 
to their food source and to the surface of the plant.  Consequently, incidental caterpillar 
mortality was much lower in 2004 than in 2003 and I was able to maintain the two 
caterpillar density treatments. 
The cotton aphid and caterpillar density treatments were applied for seven weeks 
(early July to late August) in 2003 and for nine weeks (late June to late August) in 2004.  
These periods of time roughly coincided with the reproductive period of the plants 
(square development through early boll maturation), when cotton yield is most threatened 
 
82 
by cotton aphid and caterpillar herbivory (Rosenheim et al. 1997, Mascarenhas et al. 
1999).  I sampled individual plants approximately weekly for five weeks (mid-July to late 
August) in 2003 and for seven weeks (early July to late August) in 2004.  I sampled ten 
randomly chosen plants (five from each row) in each cage in 2003 and all five plants in 
each cage in 2004.  Each week, I estimated cotton aphid density (# aphids per leaf) and 
recorded the number of fire ants and caterpillars on each plant.  I estimated cotton aphid 
density by averaging the number of aphids counted on six leaves, including the terminal 
leaf on branches four, eight, and twelve (from the top of the plant), and the mainstem leaf 
growing from the node giving rise to each of these branches.  I also quantified caterpillar 
damage to plant foliage by summing estimates of the percent leaf area consumed for each 
mainstem leaf and dividing by the total number of leaves sampled.  Finally, I quantified 
plant reproduction by recording the number of squares, flowers, and bolls on each plant, 
noting in 2004 whether the bolls were damaged or undamaged by caterpillars.  I also 
quantified plant growth in 2004 by counting the number of mainstem nodes. 
On 30 August 2003, one week after the final weekly sampling date, I collected ten 
randomly chosen plants from each cage and transported them to the lab.  Stems, leaves, 
and bolls were dried separately at 60 C for three days and then weighed to the nearest 
0.01 g to determine stem, leaf, and mean boll mass for each plant.  To better estimate 
plant reproduction in 2004, developing bolls were allowed to mature and open before 
collecting them on 20 September, four weeks after the final weekly sampling date.  I 
dried the bolls at 60 C for three days and then extracted and weighed the seedcotton 
(cotton seeds plus lint) to the nearest 0.01 g to determine total seedcotton yield for each 
plant.  Additionally, I counted the total number of seeds produced by each plant. 
 
83 
 
Greenhouse experiment 
 Because I did not manipulate the presence and absence of fire ants independently 
of cotton aphid density in the field experiments (due to a limited number of field cages), I 
implicitly assume that any reduction in caterpillar abundance and caterpillar damage to 
plants with cotton aphids is the consequence of increased fire ant predation of caterpillars 
on those plants.  This assumption is not unreasonable because fire ant abundance and 
predation of herbivores on cotton plants is strongly determined by the presence and 
abundance of cotton aphids (Kaplan and Eubanks 2002, 2005).  It is possible, however, 
that cotton aphids or their honeydew disrupt or impede foraging caterpillars, or that aphid 
herbivory alters cotton plant quality such that the plant becomes less suitable nutritionally 
for caterpillar growth and survival.  Cotton aphids, thus, could decrease caterpillar 
abundance and caterpillar damage to plants independently of fire ant predation, 
potentially reducing or even nullifying the beneficial effect of fire ants. 
I tested this hypothesis in a greenhouse experiment in which I monitored beet 
armyworm caterpillar survival in the absence of fire ants on cotton plants with and 
without cotton aphids.  Seeds from cotton of the same variety used in the 2004 field 
experiment were sown in 20.3-cm pots in a greenhouse on the Auburn University campus 
in April 2005.  Seedlings were allowed to grow until they developed their fourth set of 
true leaves at which time I transferred approximately 100 cotton aphids from a lab colony 
to each of 16 plants.  Two weeks later, I transferred five neonate beet armyworm 
caterpillars to each of the aphid-infested plants (mean ? 1 SD aphid density = 249 ? 163 
aphids per leaf) and to 15 plants without aphids.  I counted the number of caterpillars on 
 
84 
each plant every 24 hours for 10 days.  After 10 days, I collected the remaining 
caterpillars and weighed them individually to the nearest 0.01 g.  I quantified caterpillar 
damage to plants by summing estimates of the percent leaf area consumed on each leaf 
and dividing by the total number of leaves on the plant. 
 
Statistical analyses 
Because of the difficulties I had establishing the experimental treatments in 2003, 
mean cotton aphid densities varied substantially among the cages (0.1 to 24.2 aphids per 
leaf, averaged over all five sampling periods) and caterpillars densities in all cages were 
low.  Because at least some caterpillars survived the transition to the field, however, I 
could still address the first of my two questions, whether cotton aphids indirectly benefit 
cotton reproduction by increasing fire ant suppression of caterpillar herbivory.  Given the 
gradient of cotton aphid densities in the field cages, I used a series of regression analyses 
(SAS proc glm; SAS Institute, Inc. 2001) to test for: 1) a positive effect of cotton aphid 
density on fire ant abundance on cotton plants, 2) negative effects of fire ant abundance 
on the abundance of beet armyworm caterpillars and on percent caterpillar damage to 
leaves, 3) positive effects of cotton aphid density on the number of squares, flowers, and 
bolls on plants, and 4) a positive effect of cotton aphid density on mean boll mass.  I also 
regressed stem mass and leaf mass on cotton aphid density to test for effects of cotton 
aphids on vegetative growth.  Cage means averaged over all five sample dates (excepting 
mean boll mass, and stem and leaf mass, which were recorded only at harvest) were used 
in the analyses.  Values for aphid density, fire ant abundance, and caterpillar abundance 
were log(n + 1)-transformed to meet statistical assumptions. 
 
85 
 The cotton aphid and caterpillar density treatments were successfully maintained 
for the duration of the experiment in 2004.  I tested for treatment effects on cotton aphid 
density, fire ant abundance on plants, caterpillar abundance, percent caterpillar damage to 
leaves and bolls, and the number of squares, flowers, undamaged bolls, and nodes per 
plant using repeated-measures ANOVA (SAS proc mixed; SAS Institute, Inc. 2001).  I 
analyzed the number of undamaged bolls because plants quickly aborted damaged bolls.  
I tested for treatment effects on number of bolls per plant at harvest, number of seeds, and 
seedcotton mass using two-way ANOVA (SAS proc glm).  Cage means were used in all 
analyses.  Values for aphid density, fire ant abundance, caterpillar abundance, percent 
caterpillar damage to leaves and bolls, and number of squares, flowers, and bolls, were 
log(n+1)-transformed to meet statistical assumptions.  Values for number of nodes, 
number of bolls, and number of seeds were log-transformed.  Experiment-wise error rate 
was controlled in post-hoc pairwise comparisons of treatment means using sequential 
Bonferroni corrections. 
In the greenhouse experiment, I tested for a treatment effect on mean caterpillar 
survival (n surviving / 5) using repeated-measures ANOVA.  Survival data were arcsine 
square root-transformed prior to analysis.  I used two-sample t-tests (SAS proc ttest; SAS 
Institute, Inc. 2001) to compare mean caterpillar mass and mean percent caterpillar 
damage to plants between treatments.   
Means ? 1 SE are presented in the text and figures, and P-values ? 0.05 are 
considered significant in all analyses. 
 
 
 
86 
RESULTS 
2003 field experiment 
 Mean fire ant abundance on plants increased significantly with increasing mean 
cotton aphid density (F
1,26
 = 3.03, one-tailed P = 0.045, R
2
 = 0.10, b ? 1SE = 0.16 ? 0.09) 
(Fig. 1A).  Increasing fire ant abundance on plants did not result in a significant decrease 
in the mean number of beet armyworm caterpillars sampled (F
1,26
 = 0.69, one-tailed P = 
0.21, R
2
 = 0.03); however, mean percent caterpillar damage to leaves decreased 
significantly with increasing fire ant abundance on plants (F
1,26
 = 2.74, one-tailed P = 
0.05, R
2
 = 0.10, b ? 1SE = -5.69 ? 3.43) (Fig. 1B). 
 Cotton aphid density had no effect on the mean number of squares produced by 
plants (F
1,26
 = 0.92, one-tailed P = 0.17, R
2
 = 0.03).  In contrast, both the mean number of 
flowers (F
1,26
 = 2.85, one-tailed P = 0.05, R
2
 = 0.06, b ? 1SE = 0.53 ? 0.31) and the mean 
number of bolls (F
1,26
 = 3.21, one-tailed P = 0.04, R
2
 = 0.08, b ? 1SE = 0.89 ? 0.50) 
increased significantly with increasing cotton aphid density (Figs. 1C,D).  Similarly, 
mean boll mass increased significantly with increasing cotton aphid density (F
1,26
 = 5.50, 
one-tailed P = 0.01, R
2
 = 0.17, b ? 1SE = 0.47 ? 0.20) (Fig. 1E). 
 Although three of the four reproductive variables measured were associated 
positively with cotton aphid density, cotton aphids negatively affected vegetative growth.  
Both mean leaf mass (F
1,26
 = 6.32, two-tailed P = 0.02, R
2
 = 0.20, b ? 1SE = -5.16 ? 
2.05) and mean stem mass (F
1,26
 = 6.44, two-tailed P = 0.02, R
2
 = 0.20, b ? 1SE = -11.87 
? 4.68) decreased significantly with increasing cotton aphid density (Fig. 1F). 
 
 
 
87 
2004 field experiment 
 A few cotton aphids colonized plants in some of the ?aphid exclusion? cages, but 
mean aphid density was statistically indistinguishable from zero on every sampling date 
(Fig. 2A).  Mean aphid density in the ?aphids present? cages fluctuated over the seven 
sampling dates but was significantly greater than aphid density in the ?aphids excluded? 
cages on all dates (aphid density x date interaction: F
6,192
 = 3.68, P = 0.002; Fig. 2A).  
Averaged over all sampling dates, mean aphid density in the ?aphids present? cages was 
37.1 ? 2.2 aphids per leaf, and in the ?aphids excluded? cages was 5.1 ? 2.2 aphids per 
leaf (F
1,24
 = 133.47, P < 0.0001).  Mean aphid density in the ?aphids present? cages did 
not differ between the low (37.5 ? 3.1 aphids per leaf) and high (36.8 ? 3.1 aphids per 
leaf) caterpillar density treatments (P = 0.83). 
Fire ants tended cotton aphids in the ?aphids present? cages within a week after 
establishing the experimental treatments.  Fire ant abundance on plants with cotton 
aphids increased over time and was significantly greater on plants with aphids than on 
plants without aphids (i.e., plants in ?aphid exclusion? cages) on all sampling dates (aphid 
density x date interaction, Table 1) (Fig. 2B).  Averaged over all sampling dates, mean 
fire ant abundance on plants with cotton aphids was 11.5 ? 1.1 ants per plant, and on 
plants without aphids was 0.9 ? 1.1 ants per plant (aphid density effect, Table 1).   
Both the abundance of beet armyworm caterpillars on plants and percent leaf 
damage by caterpillars increased over the seven sampling dates (date effects, Table 1).  
Caterpillars were significantly less abundant on plants with cotton aphids (2.7 ? 0.5 
caterpillars per plant) than on plants without aphids (4.8 ? 0.5 caterpillars per plant; aphid 
density effect, Table 1), but this effect varied with sampling date (aphid density x date 
 
88 
interaction, Table 1) (Fig. 2C).  Caterpillars caused significantly less damage to cotton 
leaves on plants with aphids (3.2 ? 1.1% leaf damage) than on plants without aphids (7.7 
? 1.1%; aphid density effect, Table 1).  Effects of cotton aphid density on caterpillar 
abundance and caterpillar damage to leaves did not vary with caterpillar density (Table 
1).  In contrast, an effect of cotton aphid density on caterpillar damage to bolls did vary 
with caterpillar density (Table 1).  Caterpillar damage to bolls did not differ between 
plants with aphids (0.8 ? 1.7% boll damage) and without aphids (1.9 ? 1.7%) at low 
caterpillar density (P = 0.66).  At high caterpillar density, however, caterpillars damaged 
a significantly smaller proportion of bolls on plants with aphids (0.3 ? 1.7%) than on 
plants without aphids (9.2 ? 1.7%; P = 0.01).  
Averaged over all sampling dates, cotton squares, flowers, and undamaged bolls 
were significantly more abundant on plants with cotton aphids (squares: 22.4 ? 0.6; 
flowers: 2.6 ? 0.1; bolls: 9.4 ? 0.3) than on plants without aphids (squares: 18.9 ? 0.6; 
flowers: 2.0 ? 0.1; bolls: 7.8 ? 0.3; aphid density effects, Table 2).  The effect of cotton 
aphid density on number of squares varied with caterpillar density (aphid density x 
caterpillar density interaction, Table 2), however, such that mean number of squares was 
significantly greater on plants with aphids only at high caterpillar density (Fig. 3A).  
Mean number of squares was 17% greater on plants with aphids than on plants without 
aphids at high caterpillar density (P = 0.002) versus just 4% greater on plants with aphids 
at low caterpillar density (P = 0.35) (Fig 3A).  Similarly, the effect of cotton aphid 
density on number of undamaged bolls varied with caterpillar density (aphid density x 
caterpillar density x date interaction, Table 2) such that the increase in mean number of 
undamaged bolls over the season on plants with aphids was significantly greater only at 
 
89 
high caterpillar density (Fig. 3B).  On the final weekly sampling date, mean number of 
undamaged bolls was 76% greater on plants with aphids than on plants without aphids at 
high caterpillar density (P < 0.0001) versus just 11% greater on plants with aphids at low 
caterpillar density (P = 0.41) (Fig. 3B).  The effect of cotton aphid density on mean 
number of flowers per plant did not vary with caterpillar density (Table 2). 
In contrast to the three reproductive variables measured, cotton aphids negatively 
affected vegetative growth.  Averaged over all sampling dates, cotton plants with aphids 
had significantly fewer nodes (16.7 ? 0.1) than did plants without aphids (17.4 ? 0.1; 
aphid density effect, Table 2). 
At harvest, plants with aphids had significantly more cotton bolls (22.0 ? 0.8 bolls 
per plant) than did plants without aphids (19.0 ? 0.8 bolls per plant) (Table 3).  The effect 
of cotton aphids varied with caterpillar density (Table 3), however, such that mean 
number of bolls was significantly greater on plants with aphids only at high caterpillar 
density (Fig. 4A).  Mean number of bolls at harvest was 32% greater on plants with 
aphids than on plants without aphids at high caterpillar density (P = 0.002) versus just 
3% greater on plants with aphids at low caterpillar density (P = 0.80).  Plants with aphids 
also had significantly more seeds (Fig. 4B) and significantly greater seedcotton mass 
(Fig. 4C) than did plants without aphids (Table 3).  In contrast to boll number, however, 
the effects of cotton aphids on seed number and seedcotton mass did not vary with 
caterpillar density (Table 3).  Plants produced significantly more seeds at low caterpillar 
density (597.3 ? 19.4 seeds) than at high caterpillar density (535.9 ? 19.4 seeds), but 
caterpillar density had no effect on seedcotton mass (Table 3). 
 
 
90 
Greenhouse experiment 
 Mean caterpillar survival decreased over the 10-day sampling period in both the 
presence and absence of aphids (F
9,261 
= 56.20, P < 0.0001), but the average rate of 
decrease in survival was slower on plants with aphids (treatment x day interaction: F
9,261 
= 6.80, P < 0.0001).  Consequently, a greater proportion of caterpillars survived on plants 
with aphids by the final three days of sampling (Fig. 5).  As a result, percent caterpillar 
damage to leaves was significantly greater on plants with aphids (10.0 ? 1.4%) than on 
plants without aphids (6.1 ? 0.7%; unequal variance t-test: t
22.5
 = 2.42, P = 0.02).  Finally, 
although mean caterpillar mass was 16% less on plants with aphids (45.5 ? 3.3 g) than on 
plants without aphids (54.1 ? 4.7 g), it did not differ statistically between the two 
treatments (equal variance t-test: t
26
 = 1.52, P = 0.14).   
 
DISCUSSION 
 Results of the field cage experiments show that a facultative mutualism between 
cotton aphids and an invasive ant species, the red imported fire ant, can increase the 
fitness of its host plant, cotton.  In both 2003 and 2004, cotton aphids attracted fire ants 
onto plants, resulting in fire ant predation of beet armyworm caterpillars and suppression 
of caterpillar herbivory.  Decreased caterpillar damage to leaves and to developing bolls 
on plants hosting the fire ant - cotton aphid mutualism resulted in increased cotton yield 
in terms of the mean number of bolls and seeds produced per plant, and mean seedcotton 
mass.  This increase in cotton reproduction was realized in spite of direct negative effects 
of cotton aphids on plant growth (reduced stem and leaf biomass in 2003 and fewer 
mainstem nodes in 2004).  Results from the greenhouse experiment suggest that neither 
 
91 
physical interference by cotton aphids nor their effect on plant quality likely contributed 
to the reduction in caterpillar abundance and damage to plants in the field.  Caterpillar 
survival over time and caterpillar damage to leaves was actually greater in the presence of 
cotton aphids than in their absence in the greenhouse.  Combined, these results 
demonstrate that the indirect benefit to cotton plants of suppression of caterpillar 
herbivory by aphid-tending fire ants outweighed any direct cost of herbivory by cotton 
aphids themselves. 
These results are consistent with the few previous studies that have documented 
net positive effects on plants of ant-Hemipteran mutualisms involving native ants in 
terms of both reduced plant damage by non-Hemipteran herbivores (Jutsum et al. 1981, 
Khoo and Ho 1992, L?hr 1992, Floate and Whitham 1994, Dejean et al. 1997, Karhu 
1998, Van Mele and Cuc 2001, Sipura 2002, Oliveira and Del-Claro 2005) and increased 
plant growth and reproduction (Messina 1981, Whittaker and Warrington 1985, Ito and 
Higashi 1991, L?hr 1992, Sipura 2002, Suzuki et al. 2004).  One notable exception is a 
study by Buckley (1983) of the consequences of an ant-Membracid mutualism to the 
growth and reproduction of an extrafloral nectary-bearing host plant, Acacia decurrens.  
In the absence of Membracids, Iridomyrmex ants attracted onto Acacia saplings by 
extrafloral nectaries protected the plants from other herbivores, resulting in increased 
plant growth and seed set.  In the presence of Membracids, however, the ants were 
apparently attracted away from the extrafloral nectaries and spent less time patrolling the 
plant, resulting in decreased ant protection of plants and, consequently, decreased plant 
growth and seed set.  Cotton plants also bear extrafloral nectaries that attract ants, 
including fire ants.  Fire ants exhibit a much stronger numerical response to cotton aphids 
 
92 
than to extrafloral nectaries, though, resulting in comparatively greater protection to 
plants from other herbivores (Kaplan and Eubanks 2002, 2005, this study). 
 This study represents the first experimental investigation of the consequences to 
plant fitness of an ant-Hemipteran mutualism involving an invasive ant species.  
Numerical dominance, extreme aggression, and broad omnivory characterize invasive 
ants, and are apparently traits that allow invasive ants to significantly alter the abundance 
and distribution of other organisms in invaded habitats (Holway et al. 2002).  Several 
invasive ant species also possess a derived proventriculus that allows them to ingest large 
quantities of liquid food (Eisner 1957, Davidson et al. 2004), a trait that may explain the 
strong attraction of invasive ants to Hemipteran honeydew (Holway et al. 2002).  A few 
studies have investigated the effects of invasive ants as biological control agents on 
plants that host honeydew-producing Hemipterans (reviewed in Lach 2003), but none 
have tested whether plant protection by ants from non-Hemipteran herbivores was 
specifically a consequence of mutualistic interactions between invasive ants and 
Hemipterans.  Likewise, there have been no direct comparisons of the consequences to 
plants of mutualisms between Hemipterans and native versus invasive ants.  Way et al. 
(1999) observed that defoliation of the pine Pinus pinaster by the larvae of the pine 
processionary moth (Thaumetopoea pityocampa) was less severe in trees patrolled by the 
invasive Argentine ant (Linepithema humile) relative to trees patrolled by native ants.  
The authors suggested that aggregations of honeydew-producing aphids facilitated 
protection of pines by L. humile, but they did not specifically test whether the ants had 
the same effect in the absence of aphids.  Whether fire ants provide stronger protection to 
cotton plants relative to native aphid-tending ants is unknown but is very likely.  Native 
 
93 
pyramid ants (Dorymyrmex bureni) are common along field margins on my study area 
and will tend cotton aphids, but they do not show the same numerical response to cotton 
aphids as do fire ants nor are they as abundant in the interior of cotton fields (personal 
observation).  Given the potential for invasive ants to displace native Hemipteran-tending 
ants and the pervasive ecological impact of invasive ants in both natural and managed 
habitats (Holway et al. 2002, Lach 2003, Ness and Bronstein 2004), studies comparing 
the consequences to plants of interactions between Hemipterans and native versus 
invasive ants are greatly needed. 
 Our study builds on previous work by providing experimental evidence of 
conditionality in the outcome of an ant-Hemipteran mutualism to its host plant.  Messina 
(1981) and Fritz (1983) previously recognized this possibility, arguing specifically that 
the cost-to-benefit ratio of ant-Hemipteran mutualisms to plants may vary simply with the 
density of non-Hemipteran herbivores on the plant relative to Hemipteran density, as I 
show in this study.  At high caterpillar density, caterpillar damage to developing bolls 
was on average almost 8% greater on cotton plants without cotton aphids than on plants 
with aphids.  In contrast, caterpillar damage to developing bolls averaged only 1% greater 
on plants without aphids than on plants with aphids at low caterpillar density.  As a 
consequence, the mean number of bolls per plant at harvest was 32% greater on plants 
with aphids than on plants without aphids at high caterpillar density, but was no different 
between plants with and without aphids at low caterpillar density (Fig. 4A).  This 
conditionality in the benefit of cotton aphids to boll production, however, was not 
reflected in the mean number of seeds produced per plant or mean seedcotton mass.  Both 
number of seeds and seedcotton mass were significantly greater on plants with aphids, 
 
94 
but equally so at both levels of caterpillar density.  Plants at high caterpillar density may 
have compensated for the greater loss of bolls in the absence of aphids than in the 
presence of aphids by investing more resources in each boll.  Mean boll mass was 0.35 g 
(10%) greater on plants without aphids than on plants with aphids at high caterpillar 
density versus just 0.08 g (2%) greater on plants without aphids than on plants with 
aphids at low caterpillar density.  Statistically, however, the increase in mean boll mass 
on plants with aphids did not differ between the two levels of caterpillar density (F
1,24 
= 
2.07, P = 0.16). 
Sipura (2002) is the only other study that has demonstrated that the consequences 
of ant-Hemipteran mutualisms to plants can vary in different ecological contexts 
(specifically, at different levels of host plant resistance to non-Hemipteran herbivores 
[see Introduction]).  Indirect interactions between Hemipterans and other herbivores that 
are mediated by plants could also influence the effects of ant-Hemipteran mutualisms on 
plants.  Consistent with the results reported here, Rodriguez-Saona et al. (2005) found 
that aphid damage to tomato plants increased consumption of leaves and survival in S. 
exigua caterpillars.  Similarly, Stout et al. (1998) reported that S. exigua caterpillars grew 
10-50% faster when fed aphid-damaged tomato leaves than when fed aphid-free leaves.  
Because aphid herbivory may improve host plant quality for leaf-chewing herbivores, 
aphid-tending ants that are not very aggressive towards other herbivores may actually 
make plants more vulnerable to herbivory, thus increasing the net cost of aphids to plants.  
Other factors that could mediate the outcome to the host plant of ant-Hemipteran 
mutualisms include the proximity of ant colonies to host plants (Karhu 1998, Wimp and 
Whitham 2001), the abundance and efficacy of other arthropod predators (Fritz 1983), 
 
95 
variation in plant vulnerability to herbivory associated with plant phenology (Rosenheim 
et al. 1997), availability and quality of extrafloral nectar (Buckley 1983), and the effects 
of plant genotype (Floate & Whitham 1994; see also Vrieling et al. 1991) and other biotic 
(e.g., pathogen infection) and abiotic factors (e.g., nitrogen availability, shade) on host 
plant quality (Cushman 1991).  Because so many factors may influence the effects of ant-
Hemipteran mutualisms on host plants, additional studies considering the variation in 
associated costs and benefits are needed to provide greater insight into the range of 
possible outcomes and the ecological dynamics of these very common interspecific 
interactions. 
 The net effects of ant-Hemipteran mutualisms to agriculturally important 
herbaceous plants have rarely been determined empirically but are thought to be strongly 
negative because of a high potential for Hemipteran outbreaks in plant monocultures and 
because Hemipteran-vectored plant viruses can be particularly devastating to yield 
(Buckley 1987).  Cotton aphids do not transmit viruses to cotton in North America (Ebert 
and Cartwright 1997), but high densities of cotton aphids can reduce yield as a 
consequence of the loss of plant sap (Fuchs and Minzenmayer 1995, Rosenheim et al. 
1997).  I found no negative effect of aphids on cotton reproduction even at low caterpillar 
density.  These results are surprising considering that cotton aphid density averaged over 
seven weeks of plant reproduction was moderately high (~40 aphids per leaf).  There is 
no consensus on an action threshold for the treatment of cotton aphids in cotton but some 
states recommend the application of aphid pesticides when aphids reach densities of 50 
aphids per leaf.  From an applied perspective, therefore, the results of the field cage 
experiments are important because they suggest that:  (1) though typically considered 
 
96 
herbivore pests themselves, cotton aphids (at least at moderate densities) indirectly 
benefit cotton reproduction by increasing fire ant suppression of other herbivores, (2) the 
efficacy of fire ant suppression of other herbivores increases with increasing cotton aphid 
density (see also Kaplan and Eubanks 2005), and (3) the indirect benefit of cotton aphids 
is greater at higher densities of other herbivores on cotton plants.  I encourage additional 
studies of the consequences to plants of ant-Hemipteran mutualisms involving native and 
invasive ants in other annual agricultural systems to further evaluate the utility of ants as 
biological control agents of non-Hemipteran herbivores. 
 
ACKNOWLEDGMENTS 
James Bannon, Jack Dicks, Steve Nightengale, and the field staff of the E. V. 
Smith Research Center?s Row Crops Unit and Plant Breeding Unit provided exceptional 
logistical support.  I thank Lindsay Amsberry, Nick Bieser, Michael Buckman, Laura 
Cooper, Jared Gullage, Laura Mirarchi, Kevin Rose, Jennifer Nesbitt Styrsky, and Joel 
Tindle for assistance in the field.  Financial support was provided by the USDA Southern 
Region Sustainable Agriculture Research and Education (SARE) program, the Alabama 
Fire Ant Management Program, the Department of Entomology and Plant Pathology, 
Auburn University, and the Graduate School, Auburn University. 
 
LITERATURE CITED 
Bl?thgen, N., M. Verhaagh, W. Goit?a, K. Jaff?, W. Morawetz, and W. Barthlott. 2000. 
How plants shape the ant community in the Amazonian rainforest canopy: the key 
role of extrafloral nectaries and homopteran honeydew. Oecologia 125: 229-240. 
 
97 
Buckley, R. 1983. Interaction between ants and membracid bugs decreases growth and 
seed set of a host plant bearing extrafloral nectaries. Oecologia 58:132-136. 
Buckley, R. C. 1987. Interactions involving plants, homoptera, and ants. Annual Review 
of Ecology and Systematics 18:111-135. 
Carroll, C. R., and D. H. Janzen. 1973. Ecology of foraging by ants. Annual Review of 
Ecology and Systematics 4:231-257. 
Cushman, J. H. 1991 Host-plant mediation of insect mutualisms: variable outcomes in 
herbivore-ant interactions. Oikos 61:138-144. 
Cushman, J. H., S. G. Compton, C. Zachariades, A. B. Ware, R. J. C. Nefdt, and V. K. 
Rashbrook. 1998. Geographic and taxonomic distribution of a positive 
interaction: ant-tended homopterans indirectly benefit figs across southern Africa. 
Oecologia 116:373-380. 
Davidson, D. W., S. C. Cook, and R. R. Snelling. 2004. Liquid-feeding performances of 
ants (Formicidae): ecological and evolutionary implications. Oecologia 139:255-
266. 
Dejean, A., T. Bourgoin, and M. Gibernau. 1997. Ant species that protect figs against 
other ants: result of territoriality induced by a mutualistic homopteran. Ecoscience 
4:446-453. 
Delabie, J. H. C. 2001. Trophobiosis between Formicidae and Hemiptera (Sternorrhyncha 
and Auchenorrhyncha): an overview. Neotropical Entomology 30:501-516. 
Diaz, R., A. Knutson, and J. S. Bernal. 2004. Effect of the red imported fire ant on cotton 
aphid population density and predation of bollworm and beet armyworm eggs. 
Journal of Economic Entomology 97:222-229. 
 
98 
Ebert, T. A., and B. Cartwright. 1997. Biology and ecology of Aphis gossypii Glover 
(Homoptera: Aphididae). Southwestern Naturalist 22:116-153. 
Eisner, T. 1957. A comparative morphological study of the proventriculus of ants 
(Hymenoptera: Formicidae). Bulletin of the Musuem of Comparative Zoology 
116:429-490. 
Eubanks, M. D. 2001. Estimates of direct and indirect effects of red imported fire ants on 
biological control in field crops. Biological Control 21:35-43. 
Eubanks, M. D., S. A. Blackwell, C. J. Parrish, Z. D. Delamar, and H. Hull-Sanders. 
2002. Intraguild predation of beneficial arthropods by red imported fire ants in 
cotton. Environmental Entomology 31:1168-1174. 
Floate, K. D., and T. G. Whitham. 1994. Aphid-ant interaction reduces chrysomelid 
herbivory in a cottonwood hybrid zone. Oecologia 97:215-221.  
Fritz, R. S. 1983. Ant protection of a host plant?s defoliator: consequence of an ant-
membracid mutualism. Ecology 64:789-797. 
Fuchs, T. W., and R. Minzenmayer. 1995. Effect of Aphis gossypii on cotton 
development and yield in West Texas. Southwestern Entomologist 20:341-349. 
H?lldobler, B., and E. O. Wilson. 1990. The ants. Harvard University Press, Cambridge, 
Massachusetts, USA. 
Holway, D. A., L. Lach, A. V. Suarez, N. D Tsutsui, and T. J. Case. 2002. The causes 
and consequences of ant invasions. Annual Review of Ecology and Systematics 
33:181-233. 
Ito, F., and S. Higashi. 1991. An indirect mutualism between oaks and wood ants via 
aphids. Journal of Animal Ecology 60:463-470. 
 
99 
Jutsum, A. R., J. M. Cherrett, and M. Fisher. 1981. Interactions between the fauna of 
citrus trees in Trinidad and the ants Atta cephalotes and Azteca sp. Journal of 
Applied Ecology 18: 187-195. 
Kaplan, I., and M. D. Eubanks. 2002. Disruption of cotton aphid (Homoptera: Aphididae) 
? natural enemy dynamics by red imported fire ants (Hymenoptera: Formicidae). 
Environmental Entomology 31:1175-1183. 
Kaplan, I., and M. D. Eubanks. 2005. Aphids alter the community-wide impact of fire 
ants. Ecology 86:1640-1649. 
Karhu, K. J. 1998. Effects of ant exclusion during outbreaks of a defoliator and a sap-
sucker on birch. Ecological Entomology 23:185-194. 
Khoo, K. C., and C. T. Ho. 1992. The influence of Dolichoderus thoracicus 
(Hymenoptera: Formicidae) on losses due to Helopeltis theivora (Heteroptera: 
Miridae), black pod disease, and mammalian pests in cocoa in Malaysia. Bulletin 
of Entomological Research 82:485-491. 
Korzukhin M. D., S. D. Porter, L. C. Thompson, and S. Wiley. 2001. Modeling 
temperature-dependent range limits for the fire ant Solenopsis invicta 
(Hymenoptera: Formicidae) in the United States. Environmental Entomology 30: 
645-655. 
Lach, L. 2003. Invasive ants: unwanted partners in ant-plant interactions? Annals of the 
Missouri Botanical Garden 90:91-108. 
L?hr, B. 1992. The pugnacious ant, Anoplolepis custodiens (Hymenoptera: Formicidae), 
and its beneficial effect on coconut production in Tanzania. Bulletin of 
Entomological Research 82:213-218. 
 
100 
Mascarenhas, V. J., D. Cook, B. R. Leonard, E. Burris, and J. B. Graves. 1999. Late 
season beet armyworm (Lepidoptera: Noctuidae) infestations on cotton: 
defoliation, fruit damage, and yield loss. Florida Entomologist 82:218-229. 
Messina, F. J. 1981. Plant protection as a consequence of an ant-membracid mutualism: 
interactions on goldenrod (Solidago sp.). Ecology 62:1433-1440. 
Ness, J. H., and J. L. Bronstein. 2004. The effects of invasive ants on prospective ant 
mutualists. Biological Invasions 6:445-461. 
Nickerson, J. C., C. A. Rolph Kay, L. L. Buschman, and W. H. Whitcomb. 1977. The 
presence of Spissistilus festinus as a factor affecting egg predation by ants in 
soybeans. Florida Entomologist 60:193-199. 
Oliveira, P. S., and K. Del-Claro. 2005. Multitrophic interactions in a neotropical 
savanna: ant-hemipteran systems, associated insect herbivores and a host plant. 
Pages 414-438 in D. Burslem, M. Pinard, and S. Hartley, editors. Biotic 
interactions in the Tropics: their role in the maintenance of species diversity. 
Cambridge University Press, New York, New York, USA. 
Renault, C. K., L. M. Buffa, and M. A. Delfino. 2005. An aphid-ant interaction: effects 
on different trophic levels. Ecological Research 20:71-74. 
Rodriguez-Saona, C., J. A. Chalmers, S. Raj, and J. S. Thaler. 2005. Induced plant 
responses to multiple damagers: differential effects on an herbivore and its 
parasitoid. Oecologia 143:566-577. 
Rosenheim, J. A., L. R. Wilhoit, P. B. Goodell, E. E. Grafton-Cardwell, and T. F. Leigh. 
1997. Plant compensation, natural biological control, and herbivory by Aphis 
 
101 
gossypii on pre-reproductive cotton: the anatomy of a non-pest. Entomologia 
Experimentalis et Applicata 85:45-63. 
Ruberson, J. R., G. A. Herzog, W. R. Lambert, and W. J. Lewis. 1994. Management of 
the beet armyworm (Lepidoptera: Noctuidae) in cotton: role of natural enemies. 
Florida Entomologist 77:440-453. 
SAS Institute, Inc., 2001. SAS for Windows, Version 8.2. SAS Institute Inc., Cary, North 
Carolina, USA. 
Sipura, M. 2002. Contrasting effects of ants on the herbivory and growth of two willow 
species. Ecology 83:2680-2690. 
Stout, M. J., K. V. Workman, R. M. Bostock, and S. S. Duffey. 1998. Specificity of 
induced resistance in the tomato, Lycopersicon esculentum. Oecologia 113:74-81. 
Styrsky, J. D., and M. D. Eubanks. In review. Ecological consequences of ant-
Hemipteran mutualisms.  Proceedings of the Royal Society of London, B. 
Suzuki, N., K. Ogura, and N. Katayama. 2004. Efficiency of herbivore exclusion by ants 
attracted to aphids on the vetch Vicia angustifolia L. (Leguminosae). Ecological 
Research 19:275-282. 
Van Mele, P. and N. T. T. Cuc. 2001. Farmers? perceptions and practices in use of 
Dolichoderus thoracicus (Smith) (Hymenoptera: Formicidae) for biological 
control of pests of sapodilla. Biological Control 20:23-29. 
Vinson, S. B. 1997. Invasion of the red imported fire ant (Hymenoptera: Formicidae): 
spread, biology, and impact. American Entomologist 43:23-39. 
Vrieling, K., W. Smit, and E. van der Meijden. 1991. Tritrophic interactions between 
aphids (Aphis jacobaeae Schrank), ant species, Tyria jacobaeae L., and Senecio 
 
102 
jacobaea L. lead to maintenance of genetic variation in pyrrolizidine alkaloid 
concentration. Oecologia 86:177-182. 
Way, M. J. 1963. Mutualism between ants and honeydew-producing homoptera. Annual 
Review of Entomology 8:307-344. 
Way, M. J., and K. C. Khoo. 1992. Role of ants in pest management. Annual Review of 
Entomology 37:479-503. 
Way, M.J., Paiva, M.R., and Cammell, M.E. 1999 Natural biological control of the pine 
processionary moth Thaumetopoea pityocampa (Den & Schiff) by the Argentine 
ant Linepithema humile (Mayr) in Portugal. Agricultural and Forest Entomology 
1:27-31. 
Wimp, G. M., and Whitham T. G. 2001 Biodiversity consequences of predation and host 
plant hybridization on an aphid-ant mutualism. Ecology 82:440-452. 
Whittaker, J. B., and S. Warrington. 1985. An experimental field study of different levels 
of insect herbivory induced by Formica rufa predation on sycamore (Acer 
pseudoplatanus). III. Effects on tree growth. Journal of Applied Ecology 22:797-
811. 
Wojcik, D. P., C. R. Allen, R. J. Brenner, E. A. Forys, D. P. Jouvenaz, and R. S. Lutz. 
2001. Red imported fire ants: impact on biodiversity. American Entomologist 
47:16-23. 
  
103
Ta
b
l
e 1
.
  Res
u
l
t
s
 o
f
 r
e
peat
ed m
e
a
s
ur
es
 ANOV
A
 of
 t
h
e e
f
f
e
ct
s
 
o
f
 co
tto
n
 aphi
d de
n
s
i
t
y
 
a
n
d cat
e
r
p
i
l
l
a
r
 dens
i
t
y
 t
r
eat
m
e
n
t
s
 o
n
 
t
h
e 
m
e
a
n
 a
b
u
n
da
n
ce o
f
 
f
i
r
e
 an
t
s
 an
d 
b
eet
 ar
m
y
wo
r
m
 c
a
t
e
r
p
i
l
l
a
r
s
 o
n
 cotto
n
 pl
a
n
t
s
,
 an
d 
m
e
a
n
 per
c
en
t
 
l
e
a
f
 
da
m
a
ge a
n
d 
b
o
l
l
 da
m
a
ge 
by
 
cat
er
pi
ll
ar
s
 
i
n
 t
h
e 2004 f
i
e
l
d cage e
x
per
i
m
e
n
t
.
  
P
-
v
a
l
ue
s
 
f
o
r
 s
i
g
n
i
f
i
c
a
n
t
 t
r
eat
m
e
n
t
 ef
f
ect
s
 appear
 i
n
 b
o
l
d
.
 
 
 
F
i
r
e
 an
t
 ab
u
n
da
n
ce 
C
a
t
e
r
p
illa
r
 a
b
u
nda
nc
e
 
Per
c
en
t
 l
e
a
f
 da
m
a
ge 
Per
cen
t
 b
o
l
l
 da
m
a
ge
a
 
F
act
o
r
 d.
f
.
 
F 
P F P 
F P 
F P 
B
l
o
c
k 
8,
24 
  
  
 2.
22 
  
 0.
06 
  
  
  0
.
63 
  
  0
.
74 
  
  
  0
.
34 
  
 0.
94 
  
  
  1
.
22 
  
 0.
33 
A
p
hi
d de
n
s
i
t
y
 
1,
24 
  
 63.
24 
 
<
0
.
0001 
  
  11
.
87 
  
  
0
.
002 
  
  14
.
21 
  
 0.
0009 
  
  
  8
.
7
0
 
  
 0.
007 
C
a
t
e
r
p
i
l
l
a
r
 dens
i
t
y
 
1,
24 
  
  
 0.
43 
  
 0.
52 
  
  20
.
09 
  
  
0
.
0002 
  
  
  6
.
8
7
 
  
 0.
01 
  
  
  4
.
0
9
 
  
 0.
05 
Sa
m
p
li
ng dat
e
 
6,
192 
  
  
 3.
61 
  
 
0.
002 
  
  21
.
84 
  
<
0
.
0001 
  
  10
.
15 
 
<
0
.
0001 
  
  
  1
.
96 
  
 0.
15 
A
p
hi
d x
 C
a
t
e
r
p
i
l
l
a
r
 
1,
24 
  
  
 0.
04 
  
 0.
85 
  
  
  0
.
29 
  
  0
.
60 
  
  
  0
.
52 
  
 0.
48 
  
  
  5
.
40 
  
 
0.
03 
A
p
hi
d x
 Dat
e
 
6,
192 
  
  
 2.
62 
  
 
0.
02 
  
  
  2
.
6
6
 
  
  
0
.
02 
  
  
  1
.
36 
  
 0.
23 
  
  
  1
.
13 
  
 0.
33 
C
a
t
e
r
p
i
l
l
a
r
 x
 Dat
e
 
6,
192 
  
  
 1.
61 
  
 0.
15 
  
  
  2
.
30 
  
  
0
.
04 
  
  
  0
.
13 
  
 0.
99 
  
  
  0
.
44 
  
 0.
65 
A
phi
d 
x C
a
t
e
r
p
i
l
l
a
r
 x D
a
t
e
 
6,
192 
  
  
 0.
71 
  
 0.
64 
  
  
  0
.
52 
  
  0
.
80 
  
  
  0
.
31 
  
 0.
93 
  
  
  0
.
67 
  
 0.
52 
a
 A
n
a
l
y
s
i
s
 
i
n
c
l
udes
 dat
a
 f
r
o
m
 t
h
e 
f
i
na
l
 t
h
r
ee s
a
m
p
l
i
n
g dat
e
s
 o
n
l
y
 (
i
n
t
er
act
i
o
n
 t
e
r
m
s
 d.
f
.
 = 2,
64)
 b
ecaus
e
 b
o
l
l
 da
m
a
ge 
by
 
cat
er
pi
l
l
ar
s
 pr
i
o
r
 
t
o
 
t
h
es
e dat
e
s
 was
 n
e
g
l
i
g
i
b
l
e
.
  
  
104
Ta
b
l
e 2
.
  Res
u
l
t
s
 o
f
 r
e
peat
ed m
e
a
s
ur
es
 ANOV
A
 of
 t
h
e e
f
f
e
ct
s
 
o
f
 co
tto
n
 aphi
d de
n
s
i
t
y
 
a
n
d cat
er
pi
ll
ar
 dens
i
t
y
 t
r
eat
m
e
n
t
s
 o
n
 
t
h
e 
m
e
a
n
 
n
u
m
b
er
 o
f
 s
quar
e
s
,
 
f
l
o
w
er
s
,
 b
o
l
l
s
,
 an
d 
n
o
des
 
o
n
 cotto
n
 pl
a
n
t
s
 i
n
 t
h
e 2004 f
i
e
l
d cage e
x
per
i
m
e
n
t
.
 
 
P
-v
a
l
u
e
s
 f
o
r 
si
g
n
if
i
c
a
n
t
 
t
r
ea
t
m
e
n
t
 ef
f
ect
s
 appear
 
i
n
 b
o
l
d
.
 
 
 
# o
f
 
s
quar
e
s
 
# o
f
 
f
l
o
w
er
s
 
# o
f
 
u
n
da
m
a
ged 
b
o
l
l
s
# o
f
 
n
o
des
 
F
act
o
r
 d.
f
.
 
F 
P F P 
F P 
F P 
B
l
o
c
k 
8,
24 
  
  
 5.
98 
  
 0.
0003 
  
  
 1.
71 
  
 0.
15 
  
  
 1.
17 
  
 0.
36 
  
  
 6.
80 
  
 0.
0001 
A
p
hi
d de
n
s
i
t
y
 
1,
24 
  
 13.
85 
  
 
0.
0011 
  
 11.
50 
  
 
0.
002 
  
  
 7.
73 
  
 
0.
01 
  
 21.
71 
 <
0
.
0001 
C
a
t
e
r
p
i
l
l
a
r
 dens
i
t
y
 
1,
24 
  
  
 1.
22 
  
 0.
28 
  
  
 3.
40 
  
 0.
08 
  
  
 5.
02 
  
 
0.
03 
  
  
 0.
08 
  
 0.
77 
Sa
m
p
li
ng dat
e
 
6,
192 
  
 18.
49 
 <
0
.
0001 
 152.
23 
 <
0
.
0001 
 762.
94 
 <
0
.
0001 
 184.
38 
 <
0
.
0001 
A
p
hi
d x
 C
a
t
e
r
p
i
l
l
a
r
 
1,
24 
  
  
 5.
69 
  
 
0.
03 
  
  
 0.
10 
  
 0.
76 
  
  
 7.
71 
  
 
0.
01 
  
  
 0.
02 
  
 0.
89 
A
p
hi
d x
 Dat
e
 
6,
192 
  
  
 0.
41 
  
 0.
87 
  
  
 3.
80 
  
 
0.
001 
  
  
 2.
80 
  
 
0.
01 
  
  
 0.
73 
  
 0.
63 
C
a
t
e
r
p
i
l
l
a
r
 x
 Dat
e
 
6,
192 
  
  
 0.
04 
  
 0.
99 
  
  
 0.
56 
  
 0.
76 
  
  
 0.
82 
  
 0.
56 
  
  
 0.
53 
  
 0.
79 
A
phi
d 
x C
a
t
e
r
p
i
l
l
a
r
 x D
a
t
e
 
6,
192 
  
  
 0.
15 
  
 0.
99 
  
  
 1.
35 
  
 0.
24 
  
  
 3.
11 
  
 
0.
006 
  
  
 0.
56 
  
 0.
76 
 
  
105
T
a
bl
e 3.
  
R
e
s
u
l
t
s
 o
f
 AN
O
V
A o
f
 t
h
e
 e
f
f
e
c
t
s
 
o
f
 c
o
t
t
o
n a
p
h
i
d
 d
e
n
s
i
t
y
 a
n
d
 c
a
t
e
r
p
illa
r
 d
e
ns
i
t
y t
r
e
a
t
m
e
n
t
s
 
o
n
 
t
h
e
 
m
e
a
n
 
nu
mb
e
r
    
  
 
o
f
 b
o
l
l
s
 a
n
d s
eed
s
,
 an
d 
m
e
a
n
 s
e
edco
tto
n
 m
a
s
s
 h
a
r
v
es
t
e
d f
r
o
m
 co
tto
n
 pl
a
n
t
s
 
i
n
 t
h
e 2004 f
i
e
l
d cage e
x
p
e
r
i
m
e
n
t
.
 
 
P
-v
a
l
u
e
s
 f
o
r 
s
i
gn
i
f
i
c
a
n
t 
tr
e
a
tm
e
n
t e
f
f
e
c
t
s
 a
p
p
e
a
r
 i
n
 
b
o
l
d
.
 
 
 
# o
f
 
b
o
l
l
s
 
# o
f
 
s
eed
s
 
S
e
edcot
t
o
n
 
m
a
s
s
 
F
act
o
r
 d.
f
.
 
F 
P 
F P F 
P 
B
l
o
c
k 
8,
24 
  
  
  
  5
.
96 
  
  
  
  
 0.
0003 
  
  
  
  5
.
51 
  
  
  
  
 0.
0005 
  
  
  
  7
.
25 
  
  
  
 <0.
0001 
A
p
hi
d de
n
s
i
t
y
 
1,
24 
  
  
  
  6
.
72 
  
  
  
  
 0
.
0
2
 
  
  
  1
2
.2
1
 
  
  
  
  
 0.
002 
  
  
  
  3
.
5
5
 
  
  
  
  
 
0.
04 
C
a
t
e
r
p
i
l
l
a
r
 dens
i
t
y
 
1,
24 
  
  
  
  4
.
66 
  
  
  
  
 0
.
0
4
 
  
  
  
  4
.
5
8
 
  
  
  
  
 0
.
0
4
 
  
  
  
  2
.
0
2
 
  
  
  
  
 0
.
1
7
 
A
p
hi
d x
 cat
er
p
i
ll
ar
 
1,
24 
  
  
  
  4
.
93 
  
  
  
  
 0
.
0
4
 
  
  
  
  0
.
8
8
 
  
  
  
  
 0
.
3
6
 
  
  
  
  0
.
1
4
 
  
  
  
  
 0
.
7
1
 
 
 
106 
FIG. 1.  Effect of cotton aphid density (mean # per leaf) on fire ant abundance on cotton 
plants (A), effect of fire ant abundance on mean percent leaf damage by beet armyworm 
caterpillars (B), and effect of cotton aphid density on number of cotton flowers (C), 
number of cotton bolls (D), boll mass (E), and leaf and stem mass (F) in the 2003 field 
cage experiment.  Points represent field cage means.  Cotton aphid density, fire ant 
abundance, percent leaf damage, and number of cotton flowers and bolls were averaged 
over five sampling dates.  Boll, leaf, and stem mass were determined at harvest. 
 
107 
log(n+1) cotton aphid density
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
lo
g
(
n
+
1
)
 f
i
r
e
 ant
 
abun
dance
0.0
0.2
0.4
0.6
0.8
log(n+1) cotton aphid density
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
N
u
m
b
er
 of
 cot
t
on 
f
l
o
w
er
s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
log(n+1) cotton aphid density
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
me
a
n
 b
o
l
l 
ma
s
s
 (
g
)
0
1
2
3
4
5
log(n+1) fire ant abundance
0.0 0.2 0.4 0.6 0.8
%
 l
eaf
 dam
a
ge 
by cat
er
pi
l
l
as
t
0
2
4
6
8
10
12
14
log(n+1) cotton aphid density
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
num
ber
 of
 cot
t
on 
bo
l
l
s
0
1
2
3
4
5
log(n+1) cotton aphid density
0.00.20.40.60.81.01.21.4
l
eaf
 and 
s
t
em
 m
ass
 
(
g
)
0
10
20
30
40
50
60
70
leaf mass
stem mass
AB
CD
EF
 
 
108 
 
FIG. 2.  Effect of the cotton aphid manipulation (aphids included or excluded from cotton 
plants in field cages) on mean cotton aphid density (mean number per leaf) (A), mean fire 
ant abundance per plant (B), and mean beet armyworm caterpillar abundance per plant 
(C) on seven sampling dates in the 2004 field cage experiment.  Error bars represent 1 
SE.  Asterisks indicate significant differences between the two aphid treatments. 
 
109 
A
B
C
c
o
t
t
on a
phid de
ns
i
t
y
0
20
40
60
aphids present
aphids excluded
fi
r
e
 a
n
t a
bunda
nc
e
0
5
10
15
20
c
a
t
e
r
pi
ll
a
r
 a
bunda
nc
e
0
2
4
6
8
10
sampling date
7/6 7/12 7/19 7/26 8/2 8/9 8/16
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
 
 
110 
FIG. 3.  Effect of cotton aphid density (aphids included or excluded from cotton plants in 
field cages) on mean number of cotton squares (flower buds) per plant at low and high 
beet armyworm caterpillar density averaged over seven sampling dates (A), and effect of 
cotton aphid density and caterpillar density on mean number of cotton bolls undamaged 
by caterpillars per plant on seven sampling dates (B) in the 2004 field cage experiment.  
Error bars represent 1 SE.  Asterisk in (A) indicates significant difference between the 
two aphid treatments. 
 
111 
density of beet armyworm caterpillars
low high
nu
mb
e
r
 
of c
o
tto
n s
q
ua
r
e
s
0
5
10
15
20
25
30
aphids present
aphids excluded
sampling date
nu
mb
e
r
 
of c
o
tto
n b
o
l
l
s
0
5
10
15
20
25
aphids present, low caterpillar density
aphids excluded, low density
aphids present, high density
aphids excluded, high density
A
B
7/6 7/12 7/19 7/26 8/2 8/9 8/16
*
 
 
112 
FIG. 4.  Effect of cotton aphid density (aphids included or excluded from cotton plants in 
field cages) on mean number of cotton bolls per plant at harvest at low and high beet 
armyworm caterpillar density (A), and on mean number of cotton seeds per plant (B) and 
mean seedcotton mass per plant (C) at harvest averaged across the two caterpillar density 
treatments in the 2004 field cage experiment.  Error bars represent 1 SE.  Asterisks 
indicate significant differences between the two aphid treatments. 
 
113 
n
u
mb
er
 
o
f
 co
t
t
o
n
 seed
s
0
100
200
300
400
500
600
700
seed
co
t
t
o
n
 mas
s
 (
g
)
0
20
40
60
80
A
B
nu
m
b
e
r
 of
 c
o
t
t
o
n
 
bo
ll
s
0
5
10
15
20
25
30
aphids present
aphids excluded
caterpillar density
low high
C
*
*
*
 
 
114 
FIG. 5.  Mean (? 1 SE) percent beet armyworm caterpillar survival over 10 days on cotton 
plants with and without cotton aphids in the 2004 greenhouse experiment.  Asterisks 
indicate significant differences in caterpillar survival between the two aphid treatments. 
Day
%
 cat
er
pilla
r
 sur
v
iva
l
0
10
20
30
40
50
60
70
80
90
100
aphids present
aphids absent
12345678910
*
*
*
 
 
115 
CHAPTER FOUR: 
WHEN INVASIVE SPECIES MEET: PREDICTING THE ECOLOGICAL 
CONSEQUENCES OF A MUTUALISM BETWEEN SOYBEAN APHIDS 
AND RED IMPORTED FIRE ANTS 
 
Abstract.  Indirect interactions between honeydew-producing insects, other 
arthropods, and host plants mediated by direct interactions with ants are potentially 
important but often overlooked interactions in terrestrial arthropod communities.  
Previous work in cotton has shown that cotton aphids (Aphis gossypii) generate strong 
and pervasive indirect effects on other cotton arthropods and on cotton plants via their 
mutualistic interaction with an invasive predatory ant, the red imported fire ant 
(Solenopsis invicta).  In comparison, fire ants are much weaker predators in soybean and 
other crops that do not host an abundant honeydew-producing insect.  I conducted field 
and greenhouse experiments to test the hypothesis that the range expansion of an invasive 
aphid, the soybean aphid (Aphis glycines), into the range of fire ants in the southeastern 
United States will cause a dramatic increase in the community-level effects of fire ants in 
soybean.   In field experiments in which I misted soybean plants with artificial honeydew 
as a surrogate for soybean aphid honeydew, fire ants were significantly more abundant on 
honeydew-misted plants versus water-misted control plants, resulting in decreased 
 
116 
caterpillar abundance and caterpillar damage to plants.  Increased fire ant abundance on 
honeydew-misted plants had no appreciable effect, however, on other herbivore taxa or 
on non-ant predators.  Results from greenhouse experiments with soybean aphids were 
mostly consistent with the field experiments.  Significantly fewer caterpillars survived on 
soybean plants in the presence of fire ants that were tending soybean aphids than in the 
presence of fire ants alone.  Increased fire ant abundance on plants with soybean aphids 
had no effect on the survival of predatory damsel bug nymphs, but did significantly 
reduce the survival of predatory big-eyed bug nymphs.  Combined, my results suggest 
that the indirect negative effects of soybean aphids on the soybean arthropod community 
mediated by fire ants may be limited to one important group of herbivores, but that these 
effects may cascade down to benefit soybean plants. 
 
INTRODUCTION 
Over the past two decades, ecologists have increasingly recognized the prevalence 
and importance of indirect interactions in ecological communities (Wootton 1994, 
Abrams et al. 1996, Schmitz et al. 2000, Halaj and Wise 2001, Werner and Peacor 2003, 
Ohgushi 2005, van Veen et al. 2006).  Indirect interactions are interactions between two 
species mediated by a mutual interaction with a third species, as exemplified by apparent 
competition (interactions between two prey species mediated by a shared predator 
species; Holt 1977) and trophic cascades (interactions between species in two trophic 
levels [e.g., plants and predators] mediated by species in an intervening trophic level 
[e.g., herbivores]; Hairston et al. 1960, Paine 1980).  Indirect interactions strongly 
influence community structure and function (e.g., Wootton 1993, Menge 1995, Martinsen 
 
117 
et al. 1998, Wimp and Whitham 2001, Schmitz 2003) but their significance to community 
dynamics has often been overlooked. 
Many, if not most, terrestrial, plant-based arthropod communities include 
potentially important indirect interactions between honeydew-producing Hemipteran 
insects and other arthropods mediated by direct interactions with ants (Buckley 1987, 
H?lldobler & Wilson 1990,  Oliveira and Del-Claro 2005, Styrsky and Eubanks, in 
review).  Honeydew-producing Hemipterans (including aphids, scales, mealybugs, and 
treehoppers) excrete droplets of carbohydrate-rich honeydew derived from the plant sap 
on which they feed (Way 1963).  Ants are highly attracted to honeydew as a food 
resource and aggressively recruit to aggregations of honeydew-producing insects, 
defending them from predators and parasitoids (Way 1963, H?lldobler and Wilson 1990).  
Hemipteran-tending ants assiduously patrol plants and attack or harass any arthropod 
they encounter, predator or herbivore (Way 1963, Buckley 1987); thus, by causing a 
dramatic increase in the abundance and aggression of ants on plants, honeydew-
producing Hemipterans may generate potentially strong indirect effects on the abundance 
and distribution of other arthropod species utilizing the same host plant (Wimp and 
Whitham 2001, Kaplan and Eubanks 2005; reviewed in Styrsky and Eubanks, in review).  
These indirect effects may cascade down to plants if, for example, suppression of other 
herbivores by Hemipteran-tending ants causes a reduction in plant damage and a 
subsequent increase in plant fitness (Buckley 1987, Lach 2003; reviewed in Styrsky and 
Eubanks, in review).  From an applied perspective, therefore, honeydew-producing 
herbivore pests of agriculturally important plants may be better considered beneficial 
 
118 
insects in cases in which they indirectly increase biological control of more damaging 
herbivore pests (reviewed in Way and Khoo 1992, Eubanks and Styrsky 2006). 
 Previous work has shown that cotton aphids (Aphis gossypii), a common pest of 
cotton (Gossypium hirsutum) in the southeastern United States, can have strong indirect 
effects on the abundance and distribution of several species of cotton herbivores and their 
predators via a mutualistic interaction with an invasive ant, the red imported fire ant 
(Solenopsis invicta; ?fire ants? from here forward).  By manipulating densities of fire ants 
in Alabama cotton fields with naturally occurring populations of cotton aphids, Kaplan 
and Eubanks (2005) found that the mutualism between cotton aphids and fire ants 
negatively affected 27-33% of all herbivore taxa and 40-47% of all predator taxa.  In 
addition to their indirect effects on cotton arthropod community structure, cotton aphids 
can also indirectly benefit cotton plants by precipitating a trophic cascade in which 
increased predation of caterpillars by tending fire ants results in decreased caterpillar 
damage to leaves and bolls (cotton fruits), and, consequently, increased yield (Chapter 3).  
In a field experiment in which I manipulated the presence and absence of cotton aphids 
on cotton plants, plants with cotton aphids produced on average 16% more bolls, 25% 
more seeds, and 10% greater seedcotton mass than plants without aphids as a result of the 
suppression of beet armyworm (Spodoptera exigua) caterpillars by aphid-tending fire 
ants (Chapter 3). 
In comparison, the effects of fire ants as predators in other crops in Alabama 
(collards [Brassica oleracea], tomato [Lycopersicon esculentum], and soybean [Glycine 
max]) are not as strong or pervasive even though fire ant abundance on the ground in 
these three crops is roughly equivalent to that in cotton (Eubanks 2001, Eubanks et al. 
 
119 
2002, Harvey and Eubanks 2004, 2005, Cooper 2005, Styrsky and Eubanks 2006, J. D. 
Styrsky, unpublished data).  Unlike cotton, however, collard, tomato, and soybean do not 
host an abundant honeydew-producing insect, perhaps precluding strong direct and 
indirect effects of ants on other foliage arthropods and their host plants, respectively.  The 
importance of fire ants as predators in soybean may soon dramatically increase, however, 
as the range of an invasive honeydew-producing Hemipteran, the soybean aphid (Aphis 
glycines), expands south into the range of fire ants.  Soybean aphids are native herbivores 
of soybean in eastern temperate and tropical Asia and were first detected in North 
America in Wisconsin in July 2000 (Ragsdale et al. 2004).  By the end of the 2003 
growing season, soybean aphids had spread to 21 states and three Canadian provinces, 
and recent models suggest that soybean aphids will become established in all soybean 
producing regions in the United States (Venette and Ragsdale 2004). 
The spread of soybean aphids into the range of fire ants in the southeastern United 
States presents a unique opportunity to investigate the importance of indirect interactions 
between an invasive honeydew-producing insect and other arthropods mediated by an 
invasive ant.  Similar to the effect of cotton aphids in cotton, I hypothesize that soybean 
aphids will cause a dramatic shift in the importance of fire ants as predators in soybean, 
thus generating strong indirect effects on the soybean arthropod community and on 
soybean plants themselves.  I present here the results from field and greenhouse 
experiments designed to test this hypothesis.  In anticipation of the spread of soybean 
aphids into Alabama, I conducted field experiments in which I treated plants with 
artificial honeydew as a surrogate for the honeydew excreted by soybean aphids.  
Artificial honeydew has been used in many previous studies to investigate indirect effects 
 
120 
of honeydew-producing insects on other arthropods mediated by predators that are 
attracted to honeydew (e.g., Evans and England 1996; see also Hagen 1986).  I 
complemented these field experiments with greenhouse experiments in which I 
manipulated densities of soybean aphids and fire ants on plants to test whether the 
soybean aphid ? fire ant mutualism increased fire ant predation of beet armyworm 
(Spodoptera exigua) caterpillars, an important soybean herbivore, and big-eyed bugs 
(Hemiptera: Geocoridae) and damsel bugs (Hemiptera: Nabidae), two important predator 
species in soybean. 
 
METHODS 
Field experiments 
I conducted two field experiments at the E. V. Smith Research Center of the 
Auburn University Agricultural Experiment Station in Macon County, Alabama, USA 
from July through August 2004.  I conducted both experiments in a 1.75-ha field of 
soybean plants (Pioneer 96B21; 76-cm row spacings, 30 seeds per m) in an area naturally 
infested with fire ants.  In the first experiment, I established 20 plots in which I applied 
artificial honeydew (150 g sucrose / L water; Evans and England 1996) as a fine mist to 
15-m sections of two adjacent rows of plants (experimental treatment) and water to 15-m 
sections of two additional rows of plants (control treatment).  A single row of untreated 
plants separated the experimental and control rows in each plot.  Plots were separated 
from each other by 10 m and were separated from the nearest field margin by at least 15 
m.  Artificial honeydew and water were applied twice weekly for 8 weeks (from 
 
121 
beginning bloom to full seed) at a rate of 0.5 L / 15 m of row using a hand-held, 6-L 
compression sprayer. 
I sampled soybean plants for arthropods using a beat cloth in 10 plots and using a 
sweep net in the other 10 plots (Kogan and Pitrie 1980) approximately once weekly on a 
day immediately following one of the weekly applications of artificial honeydew and 
water.  Plants on either side of a 1-m
2
 cloth spread between two adjacent rows were 
beaten vigorously to dislodge all canopy arthropods.  Arthropods that fell on the beat 
cloth were quickly identified to family or lower taxonomic status and counted.  Flying 
insects cannot be sampled accurately using a beat cloth so I used sweep nets to sample 
long-legged flies (Diptera: Dolichopodidae) specifically, a very common predator in 
soybean.  The sweep net samples consisted of 10 passes of a net through the upper 50% 
of the canopy of two rows of plants.  After collecting beat cloth and sweep net samples, I 
measured the height of one haphazardly chosen plant and quantified caterpillar damage to 
leaves (beat cloth plots only) by summing estimates of the percentage leaf area consumed 
on three haphazardly chosen trifoliate leaves from the lower, middle, and upper canopy 
plus one additional leaflet and dividing by 10 (Kogan & Turnipseed 1980).  I randomly 
sampled plants from each treatment in each plot twice on each sampling date (n = 6 dates 
for beat cloth samples; n = 5 dates for sweep net samples).  Means of the two samples 
were log (n + 1)-transformed for statistical analysis. 
I tested for treatment effects on the abundance of fire ants and several herbivore 
and predator taxa on plants using repeated-measures ANCOVA (SAS proc mixed; SAS 
Institute, Inc, 2001).  I used the same model to test for treatment effects on caterpillar 
damage to plants, herbivore diversity, predator diversity, and total arthropod diversity.  
 
122 
Plant height was included as a covariate to control for any associated variation in 
arthropod abundance. 
In the second experiment, I established 10 plots in which I isolated three 
haphazardly chosen plants in a single row from the surrounding plant canopy.  
Neighboring plants were uprooted to provide approximately 15 cm of open space around 
each isolated plant, thereby preventing crawling insects from colonizing the isolated 
plants from neighboring plants.  Five-m sections of undisturbed row separated each of the 
three isolated plants in each plot, and each plot was separated by 10 m.  I applied 
artificial honeydew as a fine mist to two of the three plants and water to the third plant.  
Each plant was misted for three seconds.  To distinguish the effects of fire ants alone 
from the effects of fire ants plus all other predators, I excluded fire ants from one of the 
two plants misted with artificial honeydew by applying a thin ring of Tanglefoot? to the 
base of the plant.  Artificial honeydew and water were applied twice weekly for 8 weeks 
(from beginning bloom to full seed), and Tanglefoot was reapplied as necessary. 
 To each plant I also added five neonate beet armyworm caterpillars each week.  
Beet armyworm caterpillars were reared in the lab from commercially purchased eggs 
(Benzon Research, Carlisle, Pennsylvania, USA) and were transferred to one haphazardly 
chosen leaf on each plant a day prior to one of the weekly applications of artificial 
honeydew and water.  Beet armyworm caterpillars are highly polyphagous and feed on a 
wide variety of crop plants, including soybean. 
I visually sampled soybean plants for all arthropods approximately once weekly 
(n = 5 sample dates) on a day immediately following one of the applications of artificial 
honeydew and water.  I then measured the height of each plant and quantified caterpillar 
 
123 
damage to leaves using the same methodology as above.  Any leaflets that had been 
damaged by caterpillars prior to the experiment were removed.  Data were log (n+1)-
transformed for statistical analysis. 
I tested for treatment effects on the abundance of fire ants, several other predator 
taxa, caterpillars, and on caterpillar damage to plants using repeated measures ANCOVA 
with plant height as a covariate.  In both field experiments, experiment-wise error rate 
was controlled in post-hoc pairwise comparisons of treatment means using the Tukey-
Kramer adjustment.  Means ? 1 SE of the untransformed data are presented in the text 
and figures, and P-values ? 0.05 are considered significant in all analyses. 
 
Greenhouse experiments 
 By chance, soybean aphids appeared in very low numbers (aggregations of 10-20 
aphids on 1-3 leaves on approximately 20% of the plants visually searched in four hours) 
in several experimental soybean plots at the E. V. Smith Research Center by early 
September 2004.  Because it was too late in the season to conduct another field 
experiment, I collected approximately 300 soybean aphids to establish a greenhouse 
colony at Auburn University, Auburn, AL.  I used the progeny of these aphids in two 
greenhouse experiments in October 2004 and September 2005. 
In the first greenhouse experiment, I manipulated the presence and absence of 
soybean aphids and fire ants on individual soybean plants using a 2 x 2 randomized 
complete block design in order to test the hypothesis that a mutualistic interaction 
between soybean aphids and fire ants increases fire ant predation of caterpillars, an 
abundant group of soybean herbivores.  Seeds from the same soybean variety used in the 
 
124 
field experiments were sown in 20.3-cm pots and the seedlings (n = 28) were allowed to 
grow until they developed their fourth set of true leaves.  I transferred 50 soybean aphids 
to a single haphazardly chosen leaf on 14 individual plants and allowed them to settle and 
feed for 48 hours.  No aphids were transferred to the other 14 plants.  Half of the plants 
with soybean aphids (n = 7) and half of the plants without soybean aphids (n = 7) were 
then placed individually in plastic pans (53 x 43 x 13 cm) containing small colonies of 
approximately 2000 fire ants.  The remaining plants (n = 7 with soybean aphids and n = 7 
without soybean aphids) were placed individually in identical plastic pans without fire ant 
colonies.  The plastic pans with fire ants were lined with Fluon? to prevent the ants from 
escaping, and small twigs were propped against the rims of the pots so that fire ants could 
access the plants.  Fire ants were allowed 24 hours to discover and begin tending the 
soybean aphids. 
At the initiation of the experiment, I transferred 10 neonate beet armyworm 
caterpillars to a single haphazardly chosen leaflet (without aphids) on each plant in each 
treatment.  All 10 caterpillars were transferred to a single leaflet to mimic the natural 
tendency of neonate beet armyworms to aggregate after hatching.  Beet armyworm 
caterpillars were lab-reared from commercially purchased eggs.  I recorded the number of 
fire ants and caterpillars on each plant every 24 hours for the next four days.  Missing 
caterpillars were considered depredated.  I tested for treatment effects on fire ant 
abundance on plants (log [n+1]-transformed) and on caterpillar survival (n surviving / 10) 
using repeated measures ANOVA.  Raw data were used in the analysis of caterpillar 
survival because data transformation did not improve the fit of the model.  At the end of 
the experiment, I recorded the number of leaflets on each plant that had been damaged by 
 
125 
caterpillar feeding.  I tested for treatment effects on the percentage of caterpillar-damaged 
leaflets on each plant (n damaged leaflets / n total leaflets; log-transformed) using a two-
factor ANOVA (SAS proc glm; SAS Institute, Inc, 2001).  
The second greenhouse experiment was designed test the hypothesis that a 
mutualistic interaction between soybean aphids and fire ants increases intraguild 
predation or harassment by fire ants of big-eyed bugs and damsel bugs, two of the most 
common predator taxa in soybean fields at our study area.  Seeds from the same soybean 
variety used in the field experiments were sown in 20.3-cm pots and the seedlings (n = 
48) were allowed to grow until they developed their fourth set of true leaves.  I 
transferred 50 soybean aphids to a single haphazardly chosen leaf on 16 individual plants 
and allowed them to settle and feed for 48 hours.  I then applied artificial honeydew as a 
fine mist for 3 seconds to a second set of 16 plants.  No treatment was applied to the 
remaining 16 control plants.  Immediately thereafter, I haphazardly allocated one plant 
from each treatment to each of 16 plastic pans (53 x 43 x 13 cm) containing small 
colonies of approximately 2000 fire ants.  Each pan was thus considered a single 
replicate.  The plastic pans with fire ants were lined with Fluon? to prevent the ants from 
escaping, and small twigs were propped against the rims of the pots so that fire ants could 
access the plants.  Fire ants were allowed 24 hours to discover and begin tending the 
soybean aphids. 
At the initiation of the experiment, I divided the 16 pans into two equal groups.  I 
transferred two big-eyed bug nymphs to each plant in one group (n = 8 replicates) and 
two damsel bug nymphs to each plant in the other group (n = 8 replicates).  Third- and 
fourth-instar big-eyed bug and damsel bug nymphs were collected from agricultural 
 
126 
fields at the E. V. Smith Research Center, transferred to an environmental chamber, and 
provided only water for 24 hours prior to the experiment.  Twenty-four hours after the 
initiation of the experiment, I recorded the number of fire ants and big-eyed bug nymphs 
or damsel bug nymphs on each plant.  Missing bug nymphs were considered depredated.  
I tested for effects of the three treatments on fire ant abundance (log [n+1]-transformed) 
on plants and on big-eyed bug and damsel bug survival (n surviving / 2; arcsine square 
root-transformed) using one-factor ANOVA (SAS proc glm).  Experiment-wise error rate 
was controlled in post-hoc pairwise comparisons of treatment means using the Tukey-
Kramer adjustment.  Means ? 1 SE of the untransformed data are presented in the text 
and figures, and P-values ? 0.05 are considered significant in all analyses. 
 
RESULTS 
Field experiment 1 
Averaged over all sampling dates, fire ants were significantly more abundant on 
plants misted with artificial honeydew (4.1 ? 0.9 fire ants per sample) than on control 
plants misted with water (1.1 ? 0.9) (F
1,9
 = 23.12, P = 0.001).  Treatment effect did not 
vary with sampling date (treatment x time interaction: F
5,45 
= 2.0, P = 0.1).  In addition to 
attracting fire ants onto plants, artificial honeydew also attracted other predators.  The 
five most commonly sampled predator taxa (accounting for 97% of the total number of 
individuals sampled, excluding fire ants) were big-eyed bugs, damsel bugs, spiders, 
hooded beetles (Coleoptera: Anthicidae), and long-legged flies.  Spiders were 
significantly more abundant on plants misted with artificial honeydew than on control 
plants on three of the six sampling dates (treatment x date interaction, Table 1A; Fig. 1).  
 
127 
Averaged over all sampling dates, big-eyed bugs were 25% more abundant on plants 
misted with artificial honeydew (6.4 ? 0.3 bugs per sample) than on control plants (5.1 ? 
0.3) (Table 1A), and long-legged flies were 264% more abundant on plants misted with 
artificial honeydew (5.1 ? 0.5 flies per sample) than on control plants (1.4 ? 0.5) (Table 
1A).  Application of artificial honeydew had no effect, however, on the abundance of 
damsel bugs (honeydew-misted plants: 4.7 ? 0.3 bugs per sample; control plants: 3.9 ? 
0.3) or hooded beetles (honeydew-misted plants: 3.5 ? 0.3 beetles per sample; control 
plants: 3.0 ? 0.3) on plants (Table 1A). 
The four most commonly sampled herbivore taxa (accounting for 90% of the total 
number of individuals sampled) were Lepidopteran larvae (including velvetbean 
[Anticarsia gemmatalis], soybean looper [Pseudoplusia includens], green cloverworm 
[Plathypena scabra], and armyworm [Spodoptera spp.] caterpillars), three-cornered 
alfalfa hoppers, Spissistilus festinus [Hemiptera: Membracidae], stink bugs [Hemiptera: 
Pentatomidae], and Lygus plant bugs [Hemiptera: Miridae]).  The application of artificial 
honeydew and subsequent increase in predator abundance on plants did not affect the 
abundance of stink bugs (honeydew-misted plants: 0.8 ? 0.2 bugs per sample; control 
plants: 0.9 ? 0.2) or plant bugs (honeydew-misted plants: 0.8 ? 0.1 bugs per sample; 
control plants: 0.9 ? 0.1) in the plant canopy (Table 1B).  Averaged over all sampling 
dates, however, alfalfa hoppers were marginally significantly more abundant on plants 
misted with artificial honeydew (1.1 ? 0.1 hoppers per sample) than on control plants (0.8 
? 0.1) (Table 1B).  In contrast, there were significantly fewer caterpillars on plants misted 
with artificial honeydew than on control plants as the season progressed (treatment x date 
interaction, Table1B; Fig. 2A).  Consequently, plants misted with artificial honeydew 
 
128 
were significantly less damaged by caterpillars over the season than control plants 
(treatment x date interaction, Table 1B; Fig. 2B). 
The application of artificial honeydew and subsequent increase in fire ant 
abundance on plants did not measurably affect herbivore diversity (honeydew-misted 
plants: 5.2 ? 0.3 taxa per sample; control plants: 5.1 ? 0.1) or total arthropod diversity 
(honeydew-misted plants: 9.7 ? 0.3 taxa per sample; control plants: 9.9 ? 0.3) in the plant 
canopy (Table 2).  Averaged over all sampling dates, however, predator diversity was 
slightly but significantly reduced on plants misted with artificial honeydew (4.4 ? 0.1 
taxa per sample) relative to control plants (4.8 ? 0.1) (Table 2).   
  
Field experiment 2 
 Fire ant abundance on individual plants varied among the three treatments in the 
second field experiment (Table 3).  Fire ants were completely excluded from honeydew-
misted plants that had a ring of Tanglefoot applied around the base of the stem (Fig. 3A).  
Fire ants actively foraged on the honeydew-misted plants without the Tanglefoot ant 
barrier, however, and were significantly more abundant on those plants than on water-
misted control plants (Fig. 3A).  The abundance of all other predators combined (big-
eyed bugs, damsel bugs, long-legged flies, and spiders) on the plants also varied among 
the three treatments, but differently from that of fire ants (Table 3).  Total predator 
abundance did not differ between honeydew-misted plants with and without fire ants, and 
was significantly greater on each set of honeydew-misted plants than on control plants 
(Fig. 3B).  The only predator taxon affected individually by the three treatments, 
however, was spiders (Table 3).  Spiders were marginally significantly more abundant   
 
129 
(P = 0.05) on honeydew-misted plants without fire ants (1.0 ? 0.1 spider per plant) than 
on control plants (0.5 ? 0.1).  Spider abundance did not statistically differ between 
honeydew-misted plants with fire ants (0.9 ? 0.1) and control plants (P = 0.19), or 
between honeydew-misted plants with and without fire ants (P = 0.69). 
 Both the abundance of beet armyworm caterpillars and caterpillar damage to 
leaves varied among the three treatments (Table 3).  Caterpillars were significantly less 
abundant on honeydew-misted plants with fire ants than on water-sprayed control plants 
(P = 0.02) or on honeydew-misted plants without fire ants (P = 0.01) (Fig. 3C).  
Similarly, leaf damage by caterpillars was significantly lower on honeydew-misted plants 
with fire ants than on control plants (P = 0.002) or on honeydew-misted plants without 
fire ants (P = 0.04) (Fig. 3D).  Neither caterpillar abundance (P = 0.77) nor caterpillar 
damage to leaves (P = 0.26) measurably differed between control plants and honeydew-
misted plants without fire ants (Fig. 3C, D). 
 
Greenhouse experiment 1 
 Fire ants began tending soybean aphids within 12 hours after individual plants had 
been placed in the pans with fire ant colonies.  Averaged over the four day sampling 
period, fire ants were significantly more abundant on plants with soybean aphids (14.0 ? 
1.3 ants per plant) than on plants without soybean aphids (0.2 ? 1.4) in the two sets of 
plants that were exposed to fire ants (F
1,13 
= 86.32, P < 0.0001).  Treatment effect did not 
vary across the sampling period (no treatment x day interaction: F
3,75 
= 0.56, P = 0.64). 
 The effects of the fire ant treatment and soybean aphid treatment interacted over 
the four-day sampling period to influence beet armyworm caterpillar survival (fire ant x 
 
130 
soybean aphid x day interaction, Table 4).  Caterpillar survival remained very high over 
time on plants that were not exposed to fire ants, whether soybean aphids were present or 
absent (Fig. 4).  In contrast, caterpillar survival declined over time on plants that were 
exposed to fire ants.  Fewer caterpillars, however, survived on plants with soybean aphids 
(and tending fire ants) than on plants without soybean aphids (and far fewer fire ants) 
(Fig. 4).  By the final sampling day, caterpillar survival on plants with fire ant-tended 
soybean aphids (31 ? 4%) was half that of caterpillar survival on plants exposed to fire 
ants but without soybean aphids (62 ? 5%; P < 0.0001) (Fig. 4). 
The effects of the fire ant treatment and the soybean aphid treatment did not 
interact to influence caterpillar dispersal and feeding on plants (F
1,52 
= 0.21, P = 0.65).  
The percentage of leaflets damaged by caterpillar feeding was significantly lower on 
plants exposed to fire ants (40 ? 3%) than on plants not exposed to fire ants (72 ? 3%) 
whether soybean aphids were present or not (F
1,52 
= 64.46, P < 0.0001).  Caterpillars 
damaged a slightly lower percentage of leaflets on plants with soybean aphids (52 ? 3%) 
than on plants without soybean aphids (60 ? 3%) whether fire ants were present or not, 
but the difference was not quite statistically significant (F
1,52 
= 3.30, P = 0.07). 
 
Greenhouse experiment 2 
 Fire ants were significantly more abundant on plants with soybean aphids and on 
plants misted with artificial honeydew than on control plants (F
2,21 
= 22.16, P < 0.0001) 
in the greenhouse experiment with big-eyed bugs (Fig. 5A).  Consequently, big-eyed bug 
nymph survival was significantly greater on control plants than on plants with soybean 
aphids or on plants misted with artificial honeydew (F
2,21 
= 3.68, P = 0.04) (Fig. 5A). 
 
131 
Fire ants were also significantly more abundant on plants with soybean aphids and 
on plants misted with artificial honeydew than on control plants (F
2,21 
= 7.10, P = 0.004) 
in the greenhouse experiment with damsel bugs (Fig. 5B).  Damsel bug nymph survival, 
however, did not differ statistically among the three treatments (F
2,21 
= 0.43, P = 0.66) 
(Fig. 5B). 
 
DISCUSSION 
Previous studies of the effects of fire ants as predators in agricultural systems in 
Alabama have found that fire ants exert much stronger and more pervasive direct 
negative effects on the arthropod community in cotton than in collard, tomato, or soybean 
crops (Eubanks 2001, Eubanks et al. 2002, Harvey and Eubanks 2004, 2005, Cooper 
2005, Kaplan and Eubanks 2002, 2005, Styrsky and Eubanks 2006).  This pattern is 
attributed to the fact that cotton hosts an abundant honeydew-producing insect, the cotton 
aphid, which increases the abundance and aggressiveness of fire ants on plants, thereby 
generating strong indirect effects on other cotton arthropods and on cotton plants (Kaplan 
and Eubanks 2005, Chapter 3).  Here, I present the results of field and greenhouse 
experiments designed to determine the potential for similar indirect effects in soybean 
following the range expansion of an invasive aphid, the soybean aphid, into the range of 
fire ants.   My results suggest that soybean aphids may dramatically increase the direct 
effects of fire ants on some soybean arthropod taxa, thereby indirectly benefiting soybean 
plants. 
In the first field experiment, significantly fewer caterpillars (the most abundant 
herbivore taxon in my soybean plots) foraged on plants misted with artificial honeydew 
 
132 
than on water-misted control plants (Fig. 2A), resulting in significantly less caterpillar 
damage to leaves on honeydew-misted plants (Fig. 2B).  Whether the reduction in 
caterpillar abundance and damage was caused specifically by the greater number of fire 
ants on honeydew-sprayed plants relative to control plants is uncertain because the 
application of artificial honeydew attracted other generalist predators, including big-eyed 
bugs, long-legged flies, and spiders.  The purpose of the second field experiment, 
therefore, in which I excluded fire ants from some honeydew-misted plants using 
Tanglefoot, was to isolate the effects of fire ants on caterpillars and to determine whether 
fire ants negatively affected the non-ant predator guild.  Significantly fewer beet 
armyworm caterpillars foraged on honeydew-misted plants with fire ants than on 
honeydew-misted plants from which fire ants were excluded (Fig. 3C).  Similarly, 
honeydew-misted plants with fire ants were significantly less damaged by caterpillars 
than were honeydew-misted plants from which fire ants were excluded (Fig. 3D).  These 
results indicate that fire ants attracted to artificial honeydew had much stronger effects 
than all non-ant predators combined, and suggest that the reduced caterpillar abundance 
on honeydew-misted plants in the first field experiment was caused by fire ant predation 
specifically.  Further, combined non-ant predator abundance was no lower on honeydew-
misted plants with fire ants than on honeydew-misted plants from which fire ants were 
excluded, indicating that fire ants had no measurable effect as intraguild predators (Fig. 
3B).   
 Results of the greenhouse experiments are mostly consistent with results from the 
field experiments.  Beet armyworm caterpillar survival on plants was most reduced when 
fire ants were attracted onto plants by soybean aphids (Fig. 4), thus providing direct 
 
133 
evidence that soybean aphids can generate strong indirect effects on another important 
group of herbivores as a consequence of engaging fire ants in a mutualistic interaction.  
The percentage of soybean leaflets damaged by caterpillars in the presence of fire ants 
did not differ, however, between plants with or without soybean aphids, suggesting that 
the soybean aphid ? fire ant mutualism did not appreciably affect caterpillar dispersal on 
plants.  Inconsistent with the field data, increased fire ant abundance on plants with 
soybean aphids and on plants misted with artificial honeydew reduced the survival of big-
eyed bug nymphs compared to control plants with almost no fire ants (Fig. 5A).  In 
contrast (but consistent with the field data), increased fire ant abundance on plants with 
soybean aphids and on honeydew-misted plants had no effect on the survival of damsel 
bugs nymphs (Fig. 5B). 
Combined, the results of the field and greenhouse experiments suggest that a 
facultative mutualism between soybean aphids and fire ants will have limited effects on 
soybean arthropod community structure.  The application of artificial honeydew to plants 
strongly affected the abundance of only one important group of herbivores (caterpillars) 
via increased fire ant predation, and neither herbivore diversity nor total arthropod 
diversity differed between honeydew-misted versus control plants.  In contrast, increased 
fire ant abundance on plants misted with artificial honeydew did slightly reduce predator 
diversity, despite having no appreciable effect on the abundance of non-ant predators.  
Previous studies of the community-level consequences of mutualistic interactions 
between honeydew-producing Hemipterans and ants have reported much stronger effects.  
Fowler and MacGarvin (1985), for example, found that aphid-tending ants reduced total 
herbivore species richness in birch (Betula pubescens) by 28%.  Similarly, in an 
 
134 
experiment in which aphids were removed from cottonwood (Populus angustifolia) 
saplings, aphid-tending ants abandoned the trees resulting in an 80% increase in total 
arthropod abundance and a 57% increase in total species richness (Wimp and Whitham 
2001).  Kaplan and Eubanks (2005) did not report whether a mutualism between cotton 
aphids and fire ants affected cotton arthropod diversity in large-scale field manipulations 
of fire ant densities, but aphid-tending fire ants significantly reduced the abundance of 
27-33% of the herbivore taxa sampled and 40-47% of the predator taxa sampled. 
Low taxonomic resolution may have compromised the strength of our analyses of 
arthropod diversity in the first field experiment.  Sampling plants for arthropods using a 
beat cloth hinders identification of individuals to species because the dislodged 
arthropods must be identified and counted very quickly before they run off of the beat 
cloth.  In addition, many early-instar insects (particularly caterpillars) are too small to 
identify to species in the field with the naked eye.  Both Fowler and MacGarvin (1985) 
and Wimp and Whitham (2001) used a visual sampling technique that allowed finer 
taxonomic resolution, which may account for their ability to have documented strong 
indirect negative effects of ant-tended aphids on arthropod species richness.  Further, fire 
ants may respond more aggressively to actual aggregations of honeydew-producing 
aphids than to a sheen of artificial honeydew on the plant surface because aphid 
aggregations may be perceived by fire ants as more defendable, and aphid honeydew may 
be more valuable to fire ants as a food resource.  Artificial honeydew, therefore, may 
elicit a more anemic response in fire ants than would soybean aphid honeydew, 
weakening the overall effect of fire ants as predators of other soybean arthropods. 
 
135 
 Fire ants attracted onto plants by artificial honeydew did not have strong and 
pervasive effects on the soybean arthropod community, but they indirectly protected 
soybean plants from feeding damage caused by caterpillars.  These results are consistent 
with previous work in cotton which demonstrated that cotton aphids indirectly reduced 
caterpillar damage to plants by increasing fire ant predation of leaf- and boll-feeding 
caterpillars (Chapter 3).  Several other studies also report strong indirect effects of 
honeydew-producing Hemipterans on plant fitness via increased predation of other 
herbivores by Hemipteran-tending ants (reviewed in Styrsky and Eubanks, in review).  
Messina (1981), for instance, reported that ants that were attracted onto goldenrod plants 
(Solidago altissima) by aggregations of a honeydew-producing treehopper prevented 
plant defoliation by two leaf beetles (Chrysomelidae), resulting in greater plant growth 
and seed production relative to neighboring plants that did not host treehoppers.  
Similarly, myrmecophilous aphids on oaks (Quercus dentata) indirectly protected the 
trees from damage by acorn-boring caterpillars by attracting predaceous wood ants 
(Formica yessensis) (Ito and Higashi 1991). 
Although soybean aphids may enhance the beneficial effects of fire ants as 
predators of caterpillar pests in soybean, fire ants attracted to artificial honeydew in the 
field experiments had no measurable effect on other soybean pests, including stink bugs 
and plant bugs, and even caused a slight increase in the abundance of another Hemipteran 
pest, the three-cornered alfalfa hopper.  Alfalfa hoppers also feed on phloem sap and 
produce honeydew, but they typically occur in low densities, eliciting a much weaker 
numerical response by fire ants than do cotton aphids in cotton, for example (J. D. 
Styrsky, unpublished data).  The greater abundance of fire ants on honeydew-misted 
 
136 
plants in this study may have increased fire ant protection of alfalfa hoppers and, 
consequently, alfalfa hopper abundance.  At high densities, alfalfa hoppers can cause 
significant yield losses by feeding on the petioles and racemes during the pod-setting and 
pod-filling stages (Kogan and Turnipseed 1987); thus, soybean aphids may indirectly 
harm soybean plants by increasing ant protection of alfalfa hoppers. 
More importantly, the direct negative effects of soybean aphids themselves on 
soybean plants may outweigh any indirect benefit of increased fire ant predation of 
caterpillars.  Soybean aphid feeding can cause reduced photosynthetic capacity (Macedo 
et al. 2003), stunted growth, and reduced pod set (Nielsen and Hajek 2005), resulting in 
seed yield losses as high as 40% (DiFonzo and Hines 2002).  Soybean aphids also vector 
two important viruses of soybean plants, Soybean mosaic virus and Alfalfa mosaic virus 
(Clark and Perry 2002).   By protecting soybean aphids from predators, fire ants may 
facilitate soybean aphid outbreaks, thereby exacerbating the negative effects of soybean 
aphids on plants.  In an experimental study of the effect of mutualistic interactions 
between aphids and fire ants in tomato (Lycopersicon esculentum), for example, the 
presence of fire ants caused a 240% increase in aphid abundance on plants, resulting in a 
65% increase in the level of infection of plants by the aphid-vectored Cucumber mosaic 
virus (Cooper 2005).  The net consequence of the soybean aphid ? fire ant mutualism to 
soybean yield in any particular season and location, therefore, will depend on the 
magnitude of the indirect benefit of increased fire ant suppression of non-aphid pests 
relative to the magnitude of the direct cost of soybean aphid feeding damage.  Because 
soybean plants are actually very tolerant of damage by foliage-feeding herbivores (Kogan 
 
137 
and Turnipseed 1987), increased fire ant suppression of caterpillars in the presence of 
soybean aphids may not often outweigh damage caused by soybean aphids. 
 
ACKNOWLEDGMENTS 
Steve Nightengale and the staff of the E. V. Smith Research Center?s Plant 
Breeding Unit provided exceptional logistical support.  I thank Nick Bieser, Kevin Rice, 
and Kevin Rose for assistance in the field.  Financial support was provided by the 
Alabama Fire Ant Management Program, the Graduate School, Auburn University, and 
the Department of Entomology and Plant Pathology, Auburn University. 
 
LITERATURE CITED 
Abrams, P. A., B. A. Menge, G. G. Mittelbach, D. Spiller, and P. Yodzis. 1996. The role 
of indirect effects in food webs. Pages 371-395 in G. Polis and K. O. Winemiller, 
editors.  Food webs: integration of patterns and dynamics. Chapman and Hall, 
New York, New York, USA. 
Buckley, R. C. 1987. Interactions involving plants, homoptera, and ants. Annual Review 
of Ecology and Systematics 18:111-135. 
Clark, A. J., and K. L. Perry. 2002. Transmissibility of field isolates of soybean viruses 
by Aphis glycines. Plant Disease 86:1219-1222. 
Cooper, L. B. 2005. The potential effects of red imported fire ants (Solenopsis invicta) on 
arthropod abundance and Cucumber mosaic virus. MS Thesis, Auburn University, 
Auburn, Alabama, USA. 
 
138 
DiFonzo, C., and R. Hines. 2002. Soybean aphids in Michigan: update from the 2001 
season. MSU Extension Bulletin E-2748. Michigan State University, East 
Lansing, Michigan, USA. 
Eubanks, M. D. 2001. Estimates of direct and indirect effects of red imported fire ants on 
biological control in field crops. Biological Control 21:35-43. 
Eubanks, M. D. and J. D. Styrsky. 2006. Ant-Hemipteran mutualisms: keystone 
interactions that alter food web dynamics and influence plant reproduction. Pages 
xxx-xxx in J. Brodeur and G. Boivin, editors. Trophic and guild interactions in 
biological control. Springer, New York, New York, USA. 
Eubanks, M. D., S. A. Blackwell, C. J. Parrish, Z. D. Delamar, and H. Hull-Sanders. 
2002. Intraguild predation of beneficial arthropods by red imported fire ants in 
cotton. Environmental Entomology 31:1168-1174. 
Evans, E. W., and S. England. 1996. Indirect interactions in biological control of insects: 
pests and natural enemies in alfalfa. Ecological Applications 6:920-930. 
Fowler, S. V., and M. MacGarvin. 1985. The impact of hairy wood ants, Formica 
lugubris, on the guild structure of herbivorous insects on birch, Betula pubescens. 
Journal of Animal Ecology 54:847-855. 
Hagen, K. S. 1986. Ecosystem analysis: plant cultivars (HPR), entomophagous species 
and food supplements. Pages 151-197 in D. J. Boethel, and R. D. Eikenbarry, 
editors. Interactions of plant resistance and parasitoids and predators of insects. 
John Wiley and Sons, New York, New York, USA. 
Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. Community structure, 
population control and competition. American Naturalist 94:421-425. 
 
139 
Harvey, C. T., and M. D. Eubanks. 2004. Effect of habitat complexity on biological 
control by the red imported fire ant (Hymenoptera: Formicidae) in collards. 
Biological Control 29:348-358. 
Harvey, C. T., and M. D. Eubanks. 2005. Intraguild predation of parasitoids by 
Solenopsis invicta: a non-disruptive interaction. Entomologia Experimentalis et 
Applicata 114:127-135. 
Halaj, J., and D. H. Wise. 2001. Terrestrial trophic cascades: how much do they trickle?  
American Naturalist 157:262-281. 
H?lldobler, B., and E. O. Wilson. 1990. The ants. Harvard University Press, Cambridge, 
Massachusetts, USA. 
Holt, R. D. 1977. Predation, apparent competition, and structure of prey communities. 
Theoretical Population Biology 12:197?229. 
Ito, F., and S. Higashi. 1991. An indirect mutualism between oaks and wood ants via 
aphids. Journal of Animal Ecology 60:463-470. 
Kaplan, I., and M. D. Eubanks. 2002. Disruption of cotton aphid (Homoptera: Aphididae) 
? natural enemy dynamics by red imported fire ants (Hymenoptera: Formicidae). 
Environmental Entomology 31:1175-1183. 
Kaplan, I., and M. D. Eubanks. 2005. Aphids alter the community-wide impact of fire 
ants. Ecology 86:1640-1649. 
Kogan, M., and H. N. Pitrie Jr. 1980. General sampling methods for aboveground 
populations of soybean arthropods. Pages 30-60 in M. Kogan and D. C. Herzog, 
editors. Sampling methods in soybean entomology. Springer-Verlag, New York, 
New York, USA. 
 
140 
Kogan, M., and S. G. Turnipseed. 1980. Soybean growth and assessment of damage by 
arthropods. Pages 3-29 in M. Kogan and D. C. Herzog, editors. Sampling methods 
in soybean entomology. Springer-Verlag, New York, New York, USA. 
Kogan, M., and S. G. Turnipseed. 1987. Ecology and management of soybean 
arthropods. Annual Review of Entomology 32:507-538. 
Lach, L. 2003. Invasive ants: unwanted partners in ant-plant interactions? Annals of the 
Missouri Botanical Garden 90:91-108. 
Macedo, T. B., C. S. Bastos, L. G. Higley, K. R. Ostlie, and S. Madhaven. 2003. 
Photosynthetic responses of soybean to soybean aphid (Homoptera: Aphididae) 
injury. Journal of Economic Entomology 96:188-193. 
Martinsen, G. D., E. M. Driebe, and T. G. Whitham. 1998. Indirect interactions mediated 
by changing plant chemistry: beaver browsing benefits beetles. Ecology 79:192-
200. 
Menge, B. A. 1995. Indirect effects in marine rocky intertidal interaction webs: patterns 
and importance. Ecological Monographs 65:21?74. 
Messina, F. J. 1981. Plant protection as a consequence of an ant-membracid mutualism: 
interactions on goldenrod (Solidago sp.). Ecology 62:1433-1440. 
Nielsen, C., and A. E. Hajek. 2005. Control of invasive soybean aphid, Aphis glycines 
(Hemiptera: Aphididae), populations by existing natural enemies in New York 
state, with emphasis on entomopathogenic fungi. Environmental Entomology 
34:1036-1047. 
 
141 
Ohgushi, T. 2005. Indirect interaction webs: herbivore-induced effects through trait 
change in plants.  Annual Review of Ecology, Evolution, and Systematics 36:81-
105. 
Oliveira, P. S., and K. Del-Claro. 2005. Multitrophic interactions in a neotropical 
savanna: ant-hemipteran systems, associated insect herbivores and a host plant.  
Pages 414-438 in D. Burslem, M. Pinard, and S. Hartley, editors.  Biotic 
interactions in the Tropics: their role in the maintenance of species diversity.  
Cambridge University Press, New York, New York, USA. 
Paine, R. T. 1980. Food webs: linkage, interaction strength and community infrastructure.  
Journal of Animal Ecology 49:666-685. 
Ragsdale, D. W., D. J. Voegtlin, and R. J. O?Neil. 2004. Soybean aphid biology in North 
America. Annals of the Entomological Society of America 97:204-208. 
SAS Institute, Inc. 2001. SAS for Windows, Version 8.2. SAS Institute Inc., Cary, North 
Carolina, USA. 
Schmitz, O. J. 2003. Top predator control of plant biodiversity and productivity in an old-
field ecosystem. Ecology Letters 6:156?163. 
Schmitz, O. J., P. A. Hamb?ck, and A. P. Beckerman. 2000. Trophic cascades in 
terrestrial systems: a review of the effects of carnivore removals on plants. 
American Naturalist 155:141-153. 
Styrsky, J. D., and M. D. Eubanks. In review. Ecological consequences of ant-
Hemipteran mutualisms.  Proceedings of the Royal Society of London, B. 
 
142 
Styrsky, J. D., and M. D. Eubanks. 2006. Plant trichomes indirectly enhance tritrophic 
interactions involving a generalist predator, the red imported fire ant. Biological 
Control 36:375-384. 
van Veen, F. J. F., R. J. Morris, and H. C. J. Godfray. 2006. Apparent competition, 
quantitative food webs, and the structure of phytophagous insect communities. 
Annual Review of Entomology 51:187-208. 
Venette, R. C., and D. W. Ragsdale. 2004. Assessing the invasion by soybean aphid 
(Homoptera: Aphididae): where will it end? Annals of the Entomological Society 
of America 97:219-226. 
Way, M. J. 1963. Mutualism between ants and honeydew-producing homoptera. Annual 
Review of Entomology 8:307-344. 
Way, M.J., and K.C. Khoo. 1992. Role of ants in pest management. Annual Review of 
Entomology 37:479-503. 
Werner, E. E., and S. D. Peacor. 2003. A review of trait-mediated indirect interactions in 
ecological communities. Ecology 84:1083-1100. 
Wimp, G. M., and T. G. Whitham. 2001. Biodiversity consequences of predation and 
host plant hybridization on an aphid-ant mutualism. Ecology 82:440-452. 
Wootton, J. T. 1993. Indirect effects and habitat use in an intertidal community: 
interaction chains and interaction modifications. American Naturalist 141:71?89. 
Wootton, J. T. 1994. The nature and consequences of indirect effects in ecological 
communities. Annual Review of Ecology and Systematics 25:443-466. 
  
143 
T
a
bl
e 1.
  Res
u
l
t
s
 o
f
 r
e
peat
ed
 m
eas
ur
es
 
A
N
C
O
V
A
 o
f
 t
h
e e
f
f
e
ct
s
 
o
f
 ar
t
i
f
i
c
i
a
l
 
h
o
ne
y
d
ew o
n
 A
.
 pr
edat
or
 ab
u
n
da
n
c
e o
n
 s
o
y
b
e
a
n
 
p
l
a
n
t
s
 an
d B.
 h
e
r
b
i
v
o
r
e ab
u
n
da
nce an
d cat
er
p
i
l
l
ar
 da
m
a
ge t
o
 l
e
a
v
es
 
i
n
 
f
i
e
l
d ex
p
e
r
i
m
e
n
t
 1.
 
 S
i
g
n
i
f
i
c
a
n
t
 
t
r
ea
t
m
e
n
t
 ef
f
e
ct
s
 i
n
 
b
o
l
d.
 
A.
 P
r
ed
at
o
r
s
 
 
B
i
g-
e
y
ed b
ugs
 
Da
m
s
e
l
 b
u
g
s
 
Hoo
d
ed 
b
eet
l
e
s
 
Long
-
l
e
gge
d
 f
l
ie
s
a
 
Sp
i
d
e
r
s
 
F
act
o
r
 d.
f
.
 
F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 
T
r
ea
t
m
e
n
t
 
1,
9 
  
  
 5.
35
 (
0.
05
)
 
  
  
  3
.
60 (
0
.
09)
 
  
  3
.
70 (
0
.
09)
 
  51
.
71 (
<
0
.
0001
)
 
  21
.
51 (
0
.
001
) 
Sa
m
p
li
ng dat
e
 
5,
45 
  
  
 1.
69
 (
0
.
1
6
)
 
  
  
  2
.
23 (
0
.
07)
 
  12
.
78 (
<
0
.
0001)
 
  
  2
.
22 (
0
.
09)
 
  
  6
.
87 (
<
0
.
0001)
 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
98 
  
  
 0.
01
 (
0
.
9
4
)
 
  
  
  4
.
18 (
0
.
04)
 
  
  3
.
52 (
0
.
06)
 
  
  0
.
02 (
0
.
90)
 
  
  0
.
02 (
0
.
88)
 
T
r
e
a
tm
e
n
t x D
a
te
 
5,
45 
  
  
 0.
60
 (
0
.
7
0
)
 
  
  
  0
.
33 (
0
.
89)
 
  
  0
.
46 (
0
.
80)
 
  
  1
.
31 (
0
.
28)
 
  
  3
.
47 (
0.
0
1
) 
  
 
 
 
 
 
B
.
 He
r
b
ivor
e
s
 
 
St
i
n
k b
u
g
s
 
P
l
a
n
t
 b
ugs
 
Alf
a
l
f
a 
h
o
pper
s
 
C
a
t
e
r
p
i
l
l
a
r
s
 
C
a
t
e
r
p
illa
r
 da
m
a
g
e
 
F
act
o
r
 d.
f
.
 
F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 
T
r
ea
t
m
e
n
t
 
1,
9 
  
  
  0
.
00 (
0
.
97)
 
  
  
  0
.
42 (
0
.
53)
 
  
  
  4
.
78 (
0.
0
5
)
 
  21
.
90 (
0
.
001
)
 
  13
.
64 (
0
.
005
) 
Sa
m
p
li
ng dat
e
 
5,
45 
  
  
  1
.
62 (
0
.
17)
 
  
  
  3
.
72 (
0
.
007)
 
  
  
  2
.
33 (
0
.
05)
 
  33
.
86 (
<
0
.
0001)
 
  10
.
41 (
<
0
.
0001)
 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
98 
  
  
  1
.
63 (
0
.
20)
 
  
  
  8
.
24 (
0
.
005)
 
  
  
  0
.
83 (
0
.
37)
 
  
  0
.
00 (
0
.
96)
 
  
  0
.
07 (
0
.
80)
 
T
r
e
a
tm
e
n
t x D
a
te
 
5,
45 
  
  
  0
.
67 (
0
.
65)
 
  
  
  1
.
03 (
0
.
41)
 
  
  
  0
.
09 (
0
.
99)
 
  
  3
.
44 (
0.
0
1
)
 
  
  2
.
53 (
0.
0
4
) 
a 
Sa
m
p
l
e
d o
n
 
f
i
ve dat
e
s
 r
a
t
h
er
 
t
h
a
n
 s
i
x;
 t
h
us
,
 d.
f
.
 dat
e
 an
d 
t
r
ea
t
m
e
n
t
 x
 dat
e
 = 4
,
36,
 an
d d.
f
.
 p
l
a
n
t
 h
e
i
g
h
t
 = 1,
80
.
 
144 
Table 2.  Results of repeated measures ANCOVA of the effects of artificial honeydew on 
taxa diversity of herbivores, predators, and all arthropods on soybean plants in field 
experiment 1.  Significant treatment effects in bold. 
  Herbivores Predators All arthropods 
Factor d.f. F (P) F (P) F (P) 
Treatment 1,9         1.37 (0.27)         6.11 (0.04)         0.15 (0.71) 
Sampling date 5,45         2.47 (0.05)         3.63 (0.008)         2.27 (0.06) 
Plant height 1,98         0.27 (0.61)         9.05 (0.003)         0.59 (0.44) 
Treatment x date 
5,45         0.59 (0.71)         1.47 (0.22)         0.90 (0.49) 
  
145 
T
a
bl
e 3.
  Res
u
l
t
s
 o
f
 r
e
peat
ed
 m
eas
ur
es
 
A
N
C
O
V
A
 o
f
 t
h
e e
f
f
e
ct
s
 
o
f
 ar
t
i
f
i
c
i
a
l
 
h
o
ne
y
d
ew o
n
 t
h
e ab
u
n
da
n
c
e o
f
 
f
i
r
e
 a
n
t
s
,
 
ot
h
e
r
 
p
r
e
d
a
t
o
r
s
,
 a
n
d
 c
a
t
e
r
p
illa
r
s
,
 a
n
d
 o
n
 c
a
t
e
r
p
i
l
l
a
r
 d
a
m
a
g
e
 t
o
 s
o
ybe
a
n
 p
l
a
n
t
s
 
in 
fie
l
d
 e
x
p
e
r
i
me
nt
 2
.
 
 
 
Fi
r
e
 a
n
t
s
 
A
l
l
 o
t
h
e
r p
r
e
d
a
t
o
r
s
 
C
a
t
e
rp
i
l
l
a
rs
 
C
a
t
e
rp
i
l
l
a
r d
a
m
a
g
e
 
F
act
o
r
 d.
f
.
 
F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 
T
r
ea
t
m
e
n
t
 
2,
27 
  
  
  
  
  5
.
29 (
0.
0
1
) 
  
  
  
  6
.
8
0
 (
0
.
004
) 
  
  
  
  
  4
.
0
5
 (
0.
0
3
) 
  
  
  
  
 5
.
9
3
 (
0.
007
) 
Sa
m
p
li
ng dat
e
 
4,
106 
  
  
  
  
  1
.
52 (
0
.
10)
 
  
  
  
  8
.
44 (
<
0
.
0002)
 
  
  
  
  
  3
.
76 (
0
.
007)
 
  
  
  
 13.
35
 (
<
0.
0001
)
 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
106 
  
  
  
  
  2
.
45 (
0
.
33)
 
  
  
  32
.
00 (
<
0
.
0001)
 
  
  
  
  
  0
.
89 (
0
.
35)
 
  
  
  
  
 5.
99
 (
0
.
0
2
)
 
T
r
e
a
tm
e
n
t x D
a
te
 
8,
106 
  
  
  
  
  0
.
88 (
0
.
99)
 
  
  
  
  
 0.
49
 (
0
.
8
6
)
 
  
  
  
  
  0
.
36 (
0
.
94)
 
  
  
  
  
 0.
90
 (
0
.
6
1
)
 
  
 
 
 
 
 
 
B
i
g-
e
y
ed b
ugs
 
Da
m
s
e
l
 b
u
g
s
 
Sp
i
d
er
s
 
L
o
n
g
-
l
egged f
l
i
e
s
 
F
act
o
r
 d.
f
.
 
F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 F 
(
P
)
 
Tre
a
t
m
e
n
t
 
2
,
2
7
 
  
  
  
  
 2
.
9
9
 (0
.0
7
)
 
  
  
  
  
  1
.
4
6
 (0
.2
5
)
 
  
  
  
  3
.
3
7
 (
0.
0
5
)
 
  
  
  
  1
.
24 (
0
.
30)
 
Sa
m
p
li
ng dat
e
 
4,
106 
  
  
  
  
 5.
76
 (
0
.
0003
)
 
  
  
  
  
  1
.
25 (
0
.
30)
 
  
  
  
  8
.
81 (
<
0
.
0001)
 
  
  
  
  7
.
89 (
<
0
.
0001)
 
P
l
a
n
t
 h
e
i
g
h
t
 
1,
106 
  
  
  
  
 6.
83
 (
0
.
0
1
)
 
  
  
  
  
  0
.
80 (
0
.
37)
 
  
  
  
  1
.
68 (
0
.
20)
 
  
  
  18
.
88 (
<
0
.
0001)
 
T
r
e
a
tm
e
n
t x D
a
te
 
8,
106 
  
  
  
  
 1.
14
 (
0
.
3
4
)
 
  
  
  
  
  0
.
92 (
0
.
50)
 
  
  
  
  0
.
33 (
0
.
95)
 
  
  
  
  0
.
97 (
0
.
46)
 
 
146 
Table 4.  Results of repeated measures ANOVA on the effects of fire ants and soybean 
aphids on the survival of beet armyworm caterpillars on soybean plants in greenhouse 
experiment 1.  Significant treatment effects in bold. 
Factor F df P 
Block            0.97 13,39          0.50 
Fire ants          32.43 1,39        <0.0001 
Soybean aphids            3.84 1,39          0.06 
Day          54.80 3,149        <0.0001 
Fire ants x day          26.04   3,149        <0.0001 
Soybean aphids x day            3.43 3,149          0.02 
Fire ants x soybean aphids            3.87 1,39          0.06 
Fire ants x soybean aphids x day            3.02 3,149          0.03 
 
 
147 
FIG. 1.  Mean (? 1 SE) number of spiders per beat-cloth sample of soybean foliage misted 
with artificial honeydew or water (control treatment) in 2004.  Spiders were significantly 
more abundant on plants misted with artificial honeydew on the dates indicated with 
asterisks. 
Sampling date
Sp
id
ers p
e
r 
s
a
m
p
le
0
1
2
3
4
5
6
7
water
artificial honeydew 
7/15 7/21 7/28 8/4 8/11 8/28
*
*
*
 
 
148 
FIG. 2.  (A) Mean (? 1 SE) number of caterpillars per beat-cloth sample of soybean 
foliage, and (B) mean (? 1 SE) percent leaf damage by caterpillars on soybean plants 
misted with artificial honeydew or water (control treatment) in 2004.  Caterpillar 
abundance and caterpillar damage to leaves were significantly reduced on plants misted 
with artificial honeydew on the dates indicated with asterisks. 
Ca
t
e
r
p
il
la
r
s
 pe
r
 s
a
m
p
l
e
0
5
10
15
20
25
water
artificial honeydew 
7/15 7/21 7/28 8/4 8/11 8/28
*
*
*
A
 
Sampling date
%
 cat
erp
i
l
l
ar
 
h
e
rb
i
v
o
r
y
0
2
4
6
8
10
12
7/15 7/21 7/28 8/4 8/11 8/28
*
*
*
B
 
 
149 
FIG. 3.  Mean (? 1 SE) number of (A) fire ants per plant, (B) non-ant predators per plant, 
and (C) caterpillars per plant, and (D) mean (? 1 SE) percent leaf damage by caterpillars 
on individual soybean plants in three treatments in 2004: plants misted with water 
(control), plants misted with artificial honeydew, and plants misted with honeydew from 
which fire ants were excluded with Tanglefoot.  Within each panel, different letters above 
bars indicate significant differences among treatment means. 
  
150 
C a te rp i l l a rs  p e r p l a n t
0.
0
0.
5
1.
0
1.
5
2.
0
2.
5
3.
0
a
a
b
w
a
ter
a
rtifici
a
l
ho
neydew
honey
d
ew
+ t
a
n
g
lefoot
C
Tot al n on-an t pr e d ator s  per  plan t
0.00.51.01.52.02.5
a
b
b
w
a
ter
a
rtificial
honey
d
ew
honey
d
ew
+ tanglefoot
B
%  cate r pill ar  he r b iv or y
0.00.51.01.52.02.53.03.5
a
a
b
w
a
ter
artificial
ho
neydew
honey
d
ew
+ t
a
n
g
lefoot
D
Fir e  ants  pe r  pla n t
0.00.51.01.52.02.5
a
a
b
w
a
ter
a
rtificial
honey
d
ew
honey
d
ew
+ tanglefoot
A
tr
e
a
tm
e
n
t
tr
e
a
tm
e
n
t
 
 
151 
FIG. 4.  Mean (? 1 SE) percent survival of beet armyworm caterpillars over four days on 
soybean plants with and without soybean aphids either exposed or not exposed to fire 
ants in the 2004 greenhouse experiment.  The presence of fire ants reduced caterpillar 
survival whether soybean aphids were present or absent, but caterpillar survival was most 
reduced on plants with fire ant-tended aphids. 
time (hours)
24 48 72 96
%
 cat
er
pi
l
l
a
r
 
sur
v
i
val
0
20
40
60
80
100
no ants or aphids
soybean aphids alone
fire ants alone
ants and aphids
 
 
152 
FIG. 5.  Mean (? 1 SE) number of fire ants per plant and mean (? 1 SE) percent survival 
of (A) big-eyed bug nymphs and (B) damsel bug nymphs on individual soybean plants in 
three treatments in the 2005 greenhouse experiments: unmanipulated (control) plants, 
plants misted with artificial honeydew, and plants hosting soybean aphids.  Within each 
panel, different letters above bars of the same color indicate significant differences 
among treatment means. 
Fir
e
 ants per
 pla
n
t
0
1
2
3
4
5
6
7
%
 dam
sel bug nymph s
u
r
v
i
val
0
20
40
60
80
100
fire ants
damsel bugs
A
B
Treatment
Fir
e
 a
n
t
s
 per
 plant
0
2
4
6
8
10
%
 big-
e
yed bug nym
p
h s
u
r
v
ival
0
20
40
60
80
100
120
fire ants 
big-eyed
bugs
control artificial
honeydew
soybean
aphids
control artificial
honeydew
soybean
aphids
Treatment
a
b
b
aa
bb
bb
a
b
b
aa
aa
aa