Ants on plants: a meta-analysis of the role of ants as plant biotic defenses

PLANT-ANIMAL INTERACTIONS - ORIGINAL PAPER

Ants on plants: a meta-analysis of the role of antsas plant biotic defenses

Felix B. Rosumek Æ Fernando A. O. Silveira Æ Frederico de S. Neves ÆNewton P. de U. Barbosa Æ Livia Diniz Æ Yumi Oki Æ Flavia Pezzini ÆG. Wilson Fernandes Æ Tatiana Cornelissen

Received: 14 May 2008 / Accepted: 4 February 2009 / Published online: 7 March 2009

� Springer-Verlag 2009

Abstract We reviewed the evidence on the role of ants as

plant biotic defenses, by conducting meta-analyses for the

effects of experimental removal of ants on plant herbivory

and fitness with data pooled from 81 studies. Effects

reviewed were plant herbivory, herbivore abundance,

hemipteran abundance, predator abundance, plant biomass

and reproduction in studies where ants were experimentally

removed (n = 273 independent comparisons). Ant removal

exhibited strong effects on herbivory rates, as plants without

ants suffered almost twice as much damage and exhibited

50% more herbivores than plants with ants. Ants also

influenced several parameters of plant fitness, as plants

without ants suffered a reduction in biomass (-23.7%), leaf

production (-51.8%), and reproduction (-24.3%). Effects

were much stronger in tropical regions compared to tem-

perate ones. Tropical plants suffered almost threefold higher

herbivore damage than plants from temperate regions and

exhibited three times more herbivores. Ant removal in

tropical plants resulted in a decrease in plant fitness of about

59%, whereas in temperate plants this reduction was not

statistically significant. Ant removal effects were also more

important in obligate ant–plants (=myrmecophytes) com-

pared to plants exhibiting facultative relationships with

hemiptera or those plants with extrafloral nectaries and food

bodies. When only tropical plants were considered and the

strength of the association between ants and plants taken into

account, plants with obligate association with ants exhibited

almost four times higher herbivory compared to plants with

facultative associations with ants, but similar reductions in

plant reproduction. The removal of a single ant species

increased plant herbivory by almost three times compared to

the removal of several ant species. Altogether, these results

suggest that ants do act as plant biotic defenses, but the

effects of their presence are more pronounced in tropical

systems, especially in myrmecophytic plants.

Keywords Ant–plant interaction � Ant–plant mutualism �Formicidae � Herbivory � Indirect interactions

Introduction

Interactions between ants and plants are both ancient and

widespread (Davidson and McKey 1993; Delabie et al.

2003; Rico-Gray and Oliveira 2007). The pioneering study

of Janzen (1966) with Acacia trees and Pseudomyrmex ants

showed that ants could act as biotic defenses, protecting

plants against herbivores and parasites. In return, plants

offer benefits such as shelter and food rewards. Currently,

many plant species are known to engage in this ‘biological

warfare’, which is often recognized as a mutualistic inter-

action (Bronstein et al. 2006).

Communicated by Bernhard Stadler.

N. P. de U. Barbosa, L. Diniz, Y. Oki and F. Pezzini contributed

equally to this work and are listed in alphabetical order.

Electronic supplementary material The online version of thisarticle (doi:10.1007/s00442-009-1309-x) contains supplementarymaterial, which is available to authorized users.

F. B. Rosumek � F. A. O. Silveira � F. de S. Neves �N. P. de U. Barbosa � L. Diniz � Y. Oki � F. Pezzini �G. W. Fernandes

Instituto de Ciencias Biologicas, Universidade Federal de Minas

Gerais, Belo Horizonte, MG, Brazil

T. Cornelissen (&)

Faculdade de Ciencias Integradas do Pontal,

Universidade Federal de Uberlandia, Avenida Jose Joao Dib,

2245 Ituiutaba, MG, Brazil

e-mail: [email protected]; [email protected]

123

Oecologia (2009) 160:537–549

DOI 10.1007/s00442-009-1309-x

Ant–plant associations occur between species in most ant

subfamilies and plant species in two fern families and

innumerous phylogenetically diverse angiosperms (David-

son and McKey 1993), having evolved independently many

times (Rico-Gray and Oliveira 2007). These ant–plant

interactions are based on the array of resources provided by

plants as rewards ranging from food bodies (Fiala et al.

1989; Dutra et al. 2006), extra-floral nectaries (=EFNs,

Oliveira 1997; Rudgers 2004; Koptur 2005) or domatia

(nesting sites), the last produced by plant species called

‘‘myrmecophytes’’ (Fonseca 1994). Indirect ant–plant

interactions can also be mediated by honeydew-producing

‘‘hemipterans’’ (Order Hemiptera: Sternorryncha and

Auchenorryncha) (Compton and Robertson 1988; Crut-

singer and Sanders 2005). In return, ants deter or prey upon

insects and vertebrate herbivores and prune encroaching

vines, increasing plant fitness. However, attractor number,

identity, function, and position on the plant both directly

and indirectly influence ant recruitment and there might be

great variation in ant protection (Table 1).

Two strategies can be distinguished within defensive

ant–plant interactions. Ant–plants, or myrmecophytes, are

continuously inhabited by ants during major parts of their

life (Webber et al. 2007). Myrmecophytic plants occur in

some species of Acacia, Cecropia, Leonardoxa, Piper,

Tococa, and Macaranga, among others, and provide nesting

sites permanently inhabited by colonies of specialized ants

which protect them in a more intimate and specific associ-

ation (Davidson and McKey 1993; Heil and McKey 2003;

Rico-Gray and Oliveira 2007). Myrmecophilic plants on

the other hand are plants that, whilst providing direct

food resources that can be utilized by ants, are not regu-

larly occupied by ant colonies (Webber et al. 2007),

Myrmecophilous species offer unspecialized rewards to

attract ants, mostly in the form of extrafloral nectar or

through indirect associations with honeydew-producing

hemipterans, and gain protection from a facultative and

opportunistic ant community. Because the costs imposed

by ant-associated hemipterans may or may not be out-

weighed by the benefits that hemipteran-tending ants confer

in protecting against non-hemipteran herbivores, hemipt-

eran-mediated ant–plant associations are among the most

facultative, opportunistic, and variable of interactions

(Rico-Gray and Oliveira 2007), with a high level of uncer-

tainty for the outcomes of tending ants to host plants

(Table 1).

Since Janzen’s (1966, 1967) studies on the role of ants as

plant biotic defenses, several works have addressed how

ants protect plants by reducing herbivory rates (Bruna et al.

2004; Del-Claro et al. 2006), herbivore abundance

(Letourneau and Barbosa 1999) and richness (Crutsinger

and Sanders 2005) and ultimately, how ants increase plant

fitness by increasing plant biomass (Messina 1981), leaf

production (Freitas et al. 2000), and flower, fruit and/or

seed production (Del-Claro et al. 1996; Letourneau 1998).

Although several studies have now suggested the generality

of the pattern described by Janzen (Heil et al. 2001;

Michelangeli 2003; Dejean et al. 2006), some other studies

have found no effect of ants as plant biotic defenses (Freitas

et al. 2000) and others have even shown that ant presence

has caused negative effects to host plants (Freitas et al. 2000;

Ruhren 2003; Renault et al. 2005; Frederickson and Gordon

2007; Mooney 2007; see also Rico-Gray and Oliveira 2007).

The effects of ants as plant defenders have been

reviewed previously (Davidson and McKey 1993; Bron-

stein 1998; Heil and McKey 2003; Del-Claro 2004;

Table 1 Expected effects of ant removal under several conditions

Comparison Stronger effects Weaker effects Underlying hypothesis

Temperate 9 tropical Tropical Temperate Higher ant and herbivore diversity in the tropics, higher

herbivore specificity and ant aggressive behavior,

higher frequency of myrmecophytic plants

Single 9 several ants Single Several Exclusion of a single ant species would be more

detrimental to plant as usually single ants are

involved in obligate relationships with plants and/or

are dominant species in ant mosaics

Obligate 9 facultative association Obligate Facultative Myrmecophytic plants (=obligate associations with

ants) rely on ants for protection, whereas non-

myrmecophytic plants benefit from the relationship,

but there is no strong dependency

Ant attractors Multiple ant attractors Hemiptera Multiple ant attractors should provide most of the

energetic needs of ants, keeping colonies patrolling

plants for longer periods of time. In association with

hemipterans, however, benefits should surpass

damage caused by sucking activities

Attractor position Leaf blade Stem Attractor position on the plant might affect the types of

tissues that ants protect and the rate of recruitment

538 Oecologia (2009) 160:537–549

123

Oliveira and Freitas 2004; Bronstein et al. 2006; Rico-Gray

and Oliveira 2007), but all these reviews have been qual-

itative in nature, limiting our ability to draw general

conclusions regarding the real role of ants as plant defenses

and to what extent they increase plant fitness. In a recent

review of the consequences of interactions between ants

and honeydew-producing insects, Styrsky and Eubanks

(2007) calculated the percentage change in plant damage

and fitness from 30 studies. This approach, although

quantitative, does not take into account differences in

sample sizes and variance among studies, limiting the use

of statistical tests to examine whether ants significantly

augment their host plant fitness. Moreover, it does not

allow statistical comparisons between categories desig-

nated by the researcher. For instance, given the enormous

latitudinal differences in both diversity and productivity

between temperate and tropical habitats, it is likely that

relationships among trophic levels may also be funda-

mentally different (Dyer and Coley 2002), with different

outcomes resulting from ant–plant interactions (Table 1) in

different regions. In this study, we used meta-analysis to

review the evidence of the role of ants as plant biotic

defenses under several circumstances.

Schemske (1982) pointed out that loose or facultative

ant–plant mutualisms usually involve several species of

ants, often from several subfamilies. In contrast, myrm-

ecophytes are often associated with a single, specialized

species with often pronounced aggressive behavior (Djieto-

Lordon et al. 2007). The latter are called plant–ants (sensu

Webber et al. 2007) and their traits include colony foun-

dation in a particular host plant; a somewhat strongly

developed host specificity; host fidelity; high occupancy

rates, and intraspecific competition for host plants. In the

best-known ant–plant systems, specialized ant species differ

in their protective effects on host plants (Rico-Gray and

Oliveira 2007). Facultative ant species provide less effec-

tive or no defense in the Acacia–Pseudomyrmex system;

levels of food body production differ between facultative

and obligatorily ant-associated Macaranga species, which

may reflect variation in the degree of specialization and

intensity of interaction with Crematogaster. Finally, obli-

gate associated Azteca are considered competitively

superior to nonobligate ants in Cecropia. Therefore, spe-

cialized plant-dwelling ants are expected to provide better

defenses when compared to opportunistic species (Table 1).

Although the effects of ants as terrestrial top predators

and the effects of predators on plant damage and biomass

have been previously reviewed through meta-analyses of

trophic cascades (Halaj and Wise 2001; Schmitz et al.

2000), to the best of our knowledge, a deeper and specific

meta-analysis of the protective role of ants on plants has

yet to be done. Therefore, we aimed to use meta-analytical

methods to: (1) determine the magnitude of the effect of

ant presence on herbivory rates and parameters of plant

fitness, (2) investigate whether there are differences in the

magnitude of those effects in tropical versus temperate

regions, (3) compare the magnitude of the effects of

removal of single versus several ant species, (4) compare

the magnitude of the effects in relationship to different ant

attractors, attractor combination and number as well as

attractor position on the plant, and (5) compare the mag-

nitude of effects on obligate ant–plants (=myrmecophytes)

and facultative (=myrmecophile) relationships between

ants and host plants, in both tropical and temperate regions.

Materials and methods

The database

This meta-analytical review was based on published stud-

ies searched electronically on the Science Citation Index

Expanded (1945–2008), using ‘‘ants’’, ‘‘herbivory’’, ‘‘ant

protection’’, ‘‘ant mutualism’’, ‘‘tri-trophic interactions’’

and ‘‘ant–plant interactions’’ as keywords. We also sur-

veyed the reference list of main reviews of the role of ants

as plant biotic defenses (Delabie 2001; Oliveira and Freitas

2004; Bronstein et al. 2006; Styrsky and Eubanks 2007)

and reviews of terrestrial trophic cascades (Halaj and Wise

2001; Schmitz et al. 2000). The literature list was finally

supplemented with studies cited in the reference lists of the

articles surveyed.

Studies reviewed

To be included in our review, some criteria had to be met

by a study, such as: (1) studies that were published in

English language, (2) studies in which ants were experi-

mentally removed from the plants, creating control (with

ants) and treatment (without ants) groups, and (3) studies

that reported data with means, sample size, and a measure

of variance (standard deviation, standard error or confi-

dence intervals) for both control and treatment groups.

Studies that reported data with median or reported statis-

tical differences between control and treatment groups by

showing only F values or P values, or studies in which

sample sizes for treatment and control group were not

clear, were excluded from this meta-analytical review.

Also, studies that manipulated only hemipteran densities

(by addition or exclusion) and/or manipulated EFN (by

removal) but did not manipulate ant densities among plants

were not included in our review. Response mean values

( �Xcontrol and �Xtreatment), standard deviations (Scontrol and

Streatment) and sample size (Ncontrol and Ntreatment) were

gathered from the text, tables and/or figures from each

study included in this review. When data were available on

Oecologia (2009) 160:537–549 539

123

figures, these were digitized, and means and measurements

of variance were obtained using the software UTHSCSA

Image Tool (University of Texas, USA) after calibrating

each picture to the nearest 0.01 mm. Measurements of

variance were all converted to standard deviations of the

mean using MetaWin Statistical Calculator (Rosenberg

et al. 2000).

When data were available for several dates (several

years, months, or seasons) or several study sites, the largest

difference between control and treatment group was used

as an independent comparison. When authors used several

treatments (e.g., removal or addition of extra-floral nec-

taries, plant fertilization, removal of pollinators and/or

other predators) but did manipulate ant densities we used

data from the lesser number of inputs.

We conducted separate meta-analyses for each one of

the following nine effects of ant removal on either herbi-

vores or plant features: herbivory; herbivore, hemipteran

and predator abundance; plant biomass; leaf, flower, fruit,

and seed production. Only effects that generated at least

five independent comparisons were included in our analy-

sis. The term ‘‘herbivory’’ encompassed several variables

described by authors, such as plant damage, number of

leaves attacked, and number of leaves lost, among others.

‘‘Herbivore abundance’’ included several variables such as

infestation level, number of eggs, herbivore density, and

number of herbivores. Predators included several arthropod

taxa, but mainly spiders. The effects listed as ‘‘plant bio-

mass’’ comprised plant size, weight, height gain, and

growth. Based on information provided by the authors, we

classified studies according to the region where the study

was done (tropical vs temperate region) and according to

the type of association between ants and plants (obligate vs

facultative). Because myrmecophytes are absent in tem-

perate regions, results could be misleading due to possible

confounding effects of the inclusion of tropical myrmeco-

phytes. So, we conducted separate analyses removing all

myrmecophytes allowing for comparisons of the effects of

ant removal in tropical versus temperate species with loose

or facultative association with ants. In order to investigate

whether there are significant differences in the effects of

ant removal between myrmecophytes versus myrmeco-

philes, analyses were also run including only tropical

species.

Comparisons were also separated according to the

number of ant species (single vs several ants), where

studies that reported two or more ant species simulta-

neously were classified as ‘several’. Single ant removal

refers to cases in which only one ant species was excluded,

whereas ‘several’ refers to the exclusion of all ant species

from plants. For association type, we used data provided by

the authors or scientific literature to classify plants as

myrmecophytic or non-myrmecophytic (sensu Webber

et al. 2007). Only one study (Gaume et al. 2005) recorded a

semi-myrmecophytic plant, and inclusion or exclusion of

this category did not change our results. For ant attractor,

we recorded data as shown by authors, such as aphids,

membracids, honeydew, EFNs, etc., and later categorized

ant attractor into function and not identity, as (1) shelter

(structures produced by plants where ants build nests such

as domatia and hollow trunks), (2) food (plant-derived

rewards such as EFN, food bodies, glandular trichomes),

(3) hemipterans (insect-derived rewards such as honeydew

secretion), and (4) unknown (ant attractor was not descri-

bed on the study and/or there was no evident ant attractor).

Food bodies comprised a variety of rewards such as pearl

bodies, Mullerian bodies, Beltian bodies, and Beccarian

bodies. In many associations, not one but a combination of

rewards may be involved (Rico-Gray and Oliveira 2007),

and we also compared whether multiple attractors were

more efficient against herbivores than a single ant attractor.

For EFN-bearing plants, the position of EFN was also

categorized as leaf blade, petiole, stipule, and stem, or

reproductive structures (Rico-Gray and Oliveira 2007). To

enable evaluation of the ant effects on plant fitness

parameters under the categorical analyses, we grouped

studies that reported flower, fruit, and seed production

under the category ‘‘plant reproduction’’.

Data analysis

We used the log of the response ratio (ln ratio) to sum-

marize the effects of ant presence on both herbivores and

plant features. The response ratio is the ratio of some

measure of outcome in the experimental group to that of

the control group (Rosenberg et al. 2000) and it has the

advantage of estimating the effect as a proportionate

change resulting from experimental manipulation. We

calculated the natural log of the response ratio for each

effect studied as lr = ln �Xwithoutants=�Xcontrolð Þ (Rosenberg

et al. 2000) and effects are reported as the proportional

change from control groups (with ants). Negative per-

centage changes indicate a decrease in the plant and/or

herbivore variable compared to plants with ants and posi-

tive values indicate an increase in the effect measured due

to ant absence. To estimate the cumulative effect size

(E??) for a sample of studies addressing the same effect,

effect sizes were combined across studies using a weighted

average in which the weight for the ith study was the

reciprocal of its sampling variance (Rosenberg et al. 2000).

We used a mixed effect model of meta-analysis, in which it

is assumed that studies within a class share a common

mean effect but that there is also random variation among

studies in a class, in addition to sampling variation. Ninety-

five percent confidence limits around the effect size were

calculated and estimates of the effect sizes were considered

540 Oecologia (2009) 160:537–549

123

significant if the confidence intervals did not overlap with

zero. All analyses were conducted using MetaWin 2.1.3.4

(Rosenberg et al. 2000).

We calculated the total heterogeneity (QT) for all effects

tested and heterogeneity within (QW) and between groups

(QB). The significance of these statistics was evaluated

using a Chi-square distribution. Because our analyses were

based only on published studies, and studies that show

large and significant effects might be more likely to be

published than studies that show weak or no effects (the

‘file-drawer problem’ sensu Rosenthal 1979) we calculated

fail-safe numbers for each effect tested. Fail-safe numbers

indicate the number of non-significant, unpublished, or

missing studies that would need to be added to the sample

in order to change its results from significant to non-sig-

nificant (Rosenberg et al. 2000). As a rule of thumb, fail-

safe results are considered robust if the fail-safe number

exceeds 5k ? 10 (Moller and Jennions 2001), where k is

the number of comparisons in the analysis. We also used

funnel plots as a graphical method to assess publication

bias as, in the absence of bias, a symmetrical ‘funnel’ shape

is formed when the effect size of each study is plotted

against sample size.

Results

Qualitative results

Eighty-one studies regarding the role of ants as biotic

defenses met our criteria for inclusion in the meta-analysis,

generating 273 independent comparisons (Appendix 1).

These studies covered a wide variety of herbaceous and

woody plant taxa, including 86 host plant species in 61

genera. From these host species, two families belonged to

non-seed plant, one is a gymnosperm and 36 are angio-

sperms. Fabaceae (19 spp.) was the most common family,

followed by Euphorbiaceae (11), Salicaceae (6) and Mel-

astomataceae (6). Twenty-seven families were represented

by a single species. Species classification and nomenclature

follow APG II (2003).

Most independent comparisons came from studies con-

ducted in tropical regions (57.9%) versus 42.1% conducted

at temperate latitudes. Among myrmecophiles, 52.5% were

in temperate areas and 47.5% in tropical areas. In tropical

areas, 70.6% of the comparisons included looser association

whereas only 29.3% consisted of obligate associations.

Natural habitats (81.7%) were more studied when compared

to managed systems (18.3%). Extra-floral nectaries were the

most frequent ant attractor, accounting for 34.8% of inde-

pendent comparisons. Hemipterans represented 27.1% of

the ant attractors, followed by a combination of EFN and

hemiptera (7.3%), domatia (6.9%). and less frequent

combinations of food and shelter. Ant attractor was unde-

termined in approximately 10% of the comparisons.

Effects of a total of 20 ant genera were reviewed and ant

genus was undetermined in only 6 comparisons (2.2%).

Nearly 53% of the comparisons addressed the removal of a

particular ant species whereas in 47% of the comparisons

several ant species were excluded. When a single ant genus

was excluded, Formica was the most common genus (25.7%

of independent comparisons), followed by Pheidole (10%),

Solenopsis (10%), and Azteca (9.3%). Other ant genera

comprised 40.5% of the comparisons. Most comparisons

dealt with native ant species (95.2%), whereas only 13

comparisons (4.8%) addressed invasive ants. The insect

herbivores belonged to seven different guilds: chewers,

flower feeder, gall-inducers, miners, sap-suckers, seed pre-

dators, and stem borers, but chewers represented 67.8% of the

independent comparisons reviewed, followed by sap-suckers

(13.6%) and all the other guilds combined represented 13.0%

of all the comparisons tested. Undetermined guilds occurred

in only 5.1% of the independent comparisons.

Amongst the 273 comparisons, more than half (56.0%)

used Tanglefoot� to exclude ants, followed by the use of

insecticides (20.1%). The other 14 mechanisms of ant

exclusion accounted for 23.9% of total comparisons.

Quantitative results

We observed strong effects of ant removal on both her-

bivory rates and plant fitness. Plants without ants suffered

nearly 97% more herbivory than plants with ants

(E?? = 0.972, CI = 0.84–1.09) and exhibited 53.1%

more herbivores than control plants (E?? = 0.5,315,

CI = 0.31–0.75; Fig. 1). We also observed effects of ants

on other predators (mainly spiders), which increased by

more than 100% in plants where ants were excluded

(E?? = 1.04, CI = 0.65–1.44). As expected, a 66.2%

reduction in ‘‘Hemiptera’’ abundance was observed on ant-

excluded plants compared to ant-inhabited plants, although

this result was not significant (CI = -2.13–0.81). Ant

presence also affected plant biomass and plant reproduc-

tion. Ant exclusion increased herbivory rates and

consequently reduced plant biomass by more than 26%

(E?? = -0.264, CI = -0.41 to -0.11) and leaf produc-

tion by more than 50% (E?? = -0.518, CI = -0.76 to

–0.27). Plants without ants produced fewer fruits

(E?? = -0.381, CI = -0.71 to -0.04) and seeds

(E?? = -0.360, CI = -0.50 to -0.22). Contrary to

expectations, a trend for increased flower production was

observed on plants in which ants were removed

(E?? = 0.184, CI = -0.004–0.38), though this result was

not statistically significant.

We observed stronger effects in most parameters of

tropical regions compared to temperate ones (Fig. 2).

Oecologia (2009) 160:537–549 541

123

Plants from tropical regions experienced almost three-fold

higher damage (QB = 30.23, P \ 0.0001) and exhibited

more than three times as many herbivores (QB = 5.45,

P = 0.019) compared to temperate species. Other preda-

tors tended to be more abundant in tropical regions when

ants were excluded, but this difference was not statistically

significant (QB = 0.808, P = 0.368). The reduction in

plant fitness (production of flowers, fruits, and seeds) was

much more pronounced in tropical compared to temperate

regions. In tropical regions, ant removal resulted in a

reduction of plant fitness of nearly 59% (E?? = -0.593,

CI = -0.69 to -0.50), whereas in temperate regions this

reduction was not statistically significant (E?? = -0.054,

CI = -0.11–0.01) (QB = 108.8, P \ 0.0001).

Studies that addressed the effect of removal of a single

ant on parameters of herbivory tended to show stronger

effects compared to studies that addressed the removal of

several ant species (Fig. 3). The removal of a single ant

species increased plant herbivory by almost three times

compared to the removal of several ant species (QB =

61.93, P \ 0.0001) and those plants exhibited almost 50%

more herbivores, although this difference was not signifi-

cant (QB = 0.68, P = 0.407). For plant reproduction,

removal of several ants tended to exhibit stronger effects

on plant fitness (E?? = -0.267, CI = -0.34 to -0.19)

compared to removal of single ants (E?? = -0.1,833,

CI = -0.31 to -0.05), although this difference was not

statistically significant (QB = 1.47, P = 0.22).

Proportional change from control (with ants)-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

Herbivory

Herbivore abundance

Predator abundance

Hemiptera abundance

Plant biomass

Leaf production

Flower production

Fruit production

Seed production

(85)

(71)

(10)

(18)

(19)

(9)

(11)

(25)

(12)

Fig. 1 Effects of ant removal

on plant damage, herbivore,

predator and Hemiptera

abundance and aspects of plant

reproduction. The cumulative

effect size is reported with its

95% confidence interval.

Numbers in parentheses indicate

the number of independent

comparisons for each effect and

effects are significant if

confidence intervals do not

overlap with zero

Proportional change from control (with ants)-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5

Herbivory

Herbivore abundance

Predator abundance

Plant reproduction

(61)(25)

(35)(37)

(8)

(4)

(25) (33)

Tropical

Temperate

Fig. 2 Effects of ant removal

on plant damage, predator and

herbivore abundance and plant

reproduction according to the

region of study (tropical or

temperate). Only main effects

are presented and the effect

called ‘‘plant reproduction’’

encompasses flower, fruit and/or

seed production. The

cumulative effect size is

reported with its 95%

confidence interval and effects

are significant if confidence

intervals do not overlap with

zero

542 Oecologia (2009) 160:537–549

123

When association type was considered, only the effects

of ants on herbivory, herbivore abundance, and plant

reproduction were supported by enough data to allow cat-

egorization. Myrmecophytic plants exhibited four times

higher herbivory after ant removal than myrmecophiles

(E??obligate = 2.32, CI = 2.06–2.57; E??facultative = 0.55,

CI = 0.41–0.69). Moreover, a reduction in plant repro-

duction of approximately 55% (E??obligate = -0.56,

CI = -0.73 to -0.39) compared to 18% in non-myrmec-

ophytic plants was also observed (E??facultative = -0.177,

CI = -0.23 to -0.11). When only facultative (= looser)

associations were considered on each region, tropical

plants without ants still exhibited a stronger increase in

herbivory compared to temperate plants (E??facultative tropical

= 0.64, CI = 0.41–0.86; E??facultative temperate = 0.46,

CI = 0.23–0.69; QB = 27.31, P \ 0.05), higher herbivore

abundance (E??facultative tropical = 0.77, CI = 0.46–1.01;

E??facultative temperate = 0.25, CI = -0.07–0.585; QB =

2.25, P = 0.108), and stronger reduction in plant reproduc-

tion (E??facultative tropical = -0.62, CI = -0.71 to -0.47;

E??facultative temperate = -0.05, CI = -0.19E??0.01;

QB = 34.19, P \ 0.05). When only tropical myrmecophytes

and myrmecophiles were contrasted, ant–plants exhibited

almost 4 times higher herbivory compared to plants

with facultative associations (E??tropical obligate = 2.31,

CI = 2.02–2.60; E??tropical facultative = 0.64, CI = 0.36–

0.88; QB = 81.23, P \ 0.001), but similar reductions in

plant reproduction (E??tropical obligate = -0.51, CI = -0.79

to –0.40; E??tropical facultative = -0.46, CI = -0.74 to

-0.18; QB = 6.51, P = 0.09).

We also considered whether ant attractor types had a

significant impact on the magnitude of the effects studied.

In relationship to attractor type (= attractor identity), plants

possessing a combination of domatia and Hemiptera

exhibited the strongest effects of ant removal on herbivore

damage (E?? = 3.32), followed by plants with domatia

only (E?? = 2.42) and food bodies (E?? = 1.14). Dif-

ferent results were observed for the abundance of

herbivores, as when the category ‘‘unknown’’ was ignored,

plants hosting sap-sucking hemipterans as the sole ant

attractor exhibited the strongest effect, with a 65.7%

increase in the abundance of other non-hemipteran herbi-

vores following ant exclusion. For plant reproduction,

when to EFN-attracted ants were excluded, the reduction

on plant fitness was nearly 36% (E?? = -0.362, CI =

-0.47 to -0.24), significantly differing from the effects of

other ant attractors (QB = 28.52, P \ 0.0001). All other

ant attractors alone or in combination exhibited non-sig-

nificant effects on plant reproduction after ant exclusion

(all with confidence intervals that overlapped with zero).

When ant attractor function instead of identity was

considered, similar results were found, as plants offering

shelter and hemipterans as ant attractors exhibited the

strongest effects on herbivory rates (E?? = 3.32,

CI = 2.51–4.12), followed by shelter only (E?? = 2.43,

CI = 2.05–2.80) (Fig. 4). Weakest effects on herbivory

were observed for plants that offered only food for ants in

the form of EFNs or food bodies (E?? = 0.48, CI = 0.30–

0.66). For plant reproduction, significant effects of ant

exclusion were observed only when ants were attracted to

food rewards, with a reduction of 36% in plant fitness

compared to control plants where ants were allowed.

Multiple ant attractors exhibited strongest effects on her-

bivory (E?? = 1.49) and plant reproduction (E?? =

-0.44) compared to single ant attractors (QB = 20.61 and

QB = 10.04, respectively; P \ 0.05). When effects of

Proportional change from control (with ants)-0.5 0.0 0.5 1.0 1.5 2.0

Herbivory

Herbivore abundance

Plant reproduction

(44)(40)

(39)

(32)

(15)

(43)

single ant species

several species

Fig. 3 Effects of ant removal

on plant damage, herbivore

abundance and plant

reproduction according to the

number of ant species excluded

(either a single species or

several species). Only main

effects are presented with its

95% confidence interval. Effects

are significant if confidence

intervals do not overlap with

zero

Oecologia (2009) 160:537–549 543

123

attractor function were broken down by region, tropical

plants always suffered stronger effects of ant removal on

herbivory, herbivore abundance, and plant reproduction,

regardless of the function of the attractor (Table 2).

Because attractor position on the plant might affect the

types of tissues that ants protect, we evaluated the role of

attractor location on the magnitude of effects that allowed

a reasonable number of comparisons (n C 5). For the

analyses, we excluded comparisons in which the EFN

location was unknown or not clear in the original study

being reviewed (46 out of 273 comparisons). For

herbivory rates, strongest effects of ant protection were

observed for EFNs located on the petiole (E?? = 0.82,

CI = 0.16–1.48), followed by leaf blade (E?? = 0.66,

CI = 0.22–1.10) with significant differences among EFN

positions (QB = 42.54, P \ 0.0001). Strongest reductions

in plant fitness with ant exclusion were observed when

EFNs were located on vegetative and reproductive

structures (-33%) and on reproductive structures only

(-29%). All other categories of attractor position exhib-

ited non-significant effects on plant reproduction after ant

removal.

Proportional change from control (with ants)-1.0 0.0 1.0 2.0 3.0 4.0 5.0

Herbivory

Herbivore abundance

Predator abundance

Plant reproduction

Shelter

Shelter/Food

Hemiptera

UnknownFood

Food/Hemiptera

(5)(14)(12)(18)

(13)(22)(23)

(12)

(6)(4)

(38)

Shelter/HemipteraFig. 4 Effects of ant removal

on plant damage and

reproduction according to the

function of ant attractor in the

plant. Only main effects are

presented with its 95%

confidence interval. Effects are

significant if confidence

intervals do not overlap with

zero. For plant reproduction,

only ant attractors categorized

as food (see text for

explanation) are shown, as all

other categories exhibited very

large confidence intervals (non-

significant) that could not be

plotted on the same scale

Table 2 Effects of attractor function (see ‘‘Materials and methods’’ for explanation of categorization into function) on parameters of host plant

populations after ant removal on tropical and temperate plants

Effect/attractor function Region Number of studies E?? 95% CI QB P Fail safe number

Herbivory

Hemipteran Tropical 7 0.99 0.29 to 1.7 0.51 0.47 128

Temperate 11 0.73 0.23 to 1.2

Food Tropical 20 0.73 0.48 to 0.98 10.69 0.01* 314

Temperate 10 0.11 -0.21 to 0.45

Herbivore abundance

Hemipteran Tropical 5 0.80 -0.92 to 2.53 0.07 0.78 17

Temperate 17 0.60 -0.11 to 1.33

Food Tropical 14 0.67 0.28 to 1.05 14.47 0.001* 56

Temperate 9 0.44 -0.9 to 0.09

Plant reproduction

Food Tropical 18 -0.59 -0.72 to –0.40 11.52 0.0005* 586

Temperate 20 -0.21 -0.35 to –0.06

The only function that allowed comparisons between regions were ‘‘hemipteran’’ and ‘‘food’’, as ‘‘shelter’’ was present only in tropical plants.

The cumulative effect size is reported with its 95% confidence interval and effects are significant if confidence intervals do not overlap with zero

* Indicates statistically significant differences between tropical and temperate plants

544 Oecologia (2009) 160:537–549

123

For those plants with honeydew-producing hemipterans,

stronger effects of ant exclusion on herbivory rates were

observed when ants were associated with Coccidae

(E?? = 1.18, CI = 0.24–2.11) compared to Aphididae

(E?? = 1.00, CI = 0.47–1.54) (QB = 7.04, P \ 0.05).

Due to small sample sizes for categorization, all other

effects evaluated in this study exhibited non-significant

relationships with ant exclusion according to hemipteran

family.

Assessment of publication bias

Fail-safe numbers for effects of ant removal on herbivory

rates and herbivore abundance were large (8,606 and 551

studies, respectively) relative to the number of independent

comparisons included in the meta-analysis (85 and 71,

respectively). For the effect grouped into ‘‘plant repro-

duction’’, the fail-safe number was 1,163 studies,

indicating the strength of the results found. Scatter plots of

effects size against sample size of all data exhibited a

typical funnel shape (figure not shown), indicating that

studies with low precision—generally with small sample

sizes—show a large dispersal of effect sizes around the true

effect, whereas those with large sample sizes have an effect

size close to the true value. Our results suggest that there

was little publication bias in the studies included in this

meta-analytical review.

Discussion

This meta-analytical review provides strong support for

ant–plant symbioses, although it also reveals the diversity

of ant–plant interactions with differential effects of ant

removal according to latitude, type of association between

ants and plants, number of ants involved in these associa-

tions, and type of ant attractor. The evidence presented here

reinforces the role of ants as plant biotic defenses, with

increased herbivory rates, herbivore density, and reduced

plant fitness following ant exclusion. Effects of ant

removal, however, were not homogeneous on plant

parameters investigated. Ant effects on herbivory rates

were stronger than their effects on insect herbivore density

and/or predator abundance. Ant removal increased her-

bivory by more than 95%, whereas increased herbivore

abundance by only 50%. Quantification of ant effects on

herbivory loads suffered by plants might be actually easier

to perform than quantification of herbivore abundance after

ant exclusion, as herbivory levels might be investigated

only once during a course of a study (a snapshot of effects

of ant exclusion in a season, such as number of chewed

leaves or feeding holes), whereas data collection on her-

bivore abundance encompasses direct observations and/or

comparatively higher sampling efforts. Moreover, herbi-

vores might increase residence time on plants without ants

(Suzuki et al. 2004) therefore increasing the amount of

damage on the plants they feed upon and contributing to

stronger effects of ant removal on herbivory compared to

effects on insect abundance.

The effects of ant–plant or ant–plant-hemipteran asso-

ciations on herbivory is also species-specific, with different

outcomes in partners’ fitness (Heil and McKey 2003). This

variation in strength among species is dependent upon the

vulnerability of herbivores to ant predation and/or avoid-

ance, and can lead to changes in the overall structure of

arthropod communities in the presence of ants (Fowler and

MacGarvin 1985). Styrsky and Eubanks (2007) have

shown that species richness of other sucking, non-honey-

dew-producing herbivores was reduced by 28% after ant

exclusion, whereas species richness of leaf-chewing cat-

erpillars was increased by 69% on branches without ants

compared to branches with ants. In contrast, species rich-

ness of leaf-mining caterpillars, a guild of herbivores that is

protected against ant predation, was actually 44% greater

on the trees with hemipteran-tending ants, presumably

because the ants indirectly protect the concealed caterpil-

lars from other predators (Fowler and MacGarvin 1985).

Our review has shown that, although herbivores studied

belonged to seven different guilds, chewers such as cater-

pillars and folivorous beetles accounted for 68% of the

independent comparisons reviewed, impairing analyses

regarding the effects of ant removal on herbivory rates

according to herbivore feeding mode. In addition, ant

behavior may also affect the outcome of the interaction.

Ants vary greatly in their behavior towards herbivores

(Bronstein 1998; Michelangeli 2003) and in the way they

interact with insects, therefore differing in their protective

effect on host plants (Fraser et al. 2001). The function

of ant attractant and the ultimate effects on a plant

thus depend on the array of species visiting the attractant

(Cuautle and Rico-Gray 2003).

Ecologically dominant ants have been shown to alter

arthropod communities, acting as important predators,

mutualists, competitors, and prey (Gibb 2003). The fact

that hemipteran-tending ants reduce the survival and

abundance and alter the spatial distribution of hemipterans’

natural enemies is extensively documented (Del-Claro and

Oliveira 2000; Renault et al. 2005). Our results have shown

that ant exclusion caused an increase in predator abundance

and a tendency to a decrease in the abundance of hemipt-

erans. In the 14 comparisons of changes in the density of

other predators from nine independent studies, the majority

of predators studied were generalist predators, such as

spiders. Some authors have argued that effects of ants on

other predators tend to be less consistent than effects of

ants on herbivores, with some predator taxa responding to

Oecologia (2009) 160:537–549 545

123

ant exclusion (Sipura 2002) and others not (Karhu 1998;

Gibb 2003; Offenberg et al. 2005). Our results for ant

effects on predators come from a relatively small number

of comparisons, but they do show strong effects of ant

removal on predator abundance in terrestrial communities

which, in turn, have the potential to affect the composition,

diversity, and abundance of the herbivore community on

those plants. Increase in the abundance of other predators

with ant exclusion might be explained by increased

vulnerability of ant-tended hemiptera in the absence of

ants, making them easy targets for spiders (Del-Claro and

Oliveira 2000), beetles (James et al. 1999), and syrphid

flies (Sipura 2002). Ant exclusion allows the attack of

natural enemies of hemipterans, thereby decreasing their

abundance (by 66% in our review, although not significant)

and supporting the evidence for the strong trophobiosis

between Formicidae and hemipterans (see Delabie 2001).

Some of the most central functions enabling plant sur-

vival and reproduction depend on mutualisms (Heil 2008),

and many plants rely on the third trophic level in order to

get protection from the second trophic level. Beneficial

effects of ants on plants arise when ants are capable of

reducing herbivore numbers, herbivory levels, or both,

thereby decreasing plant damage and increasing plant fit-

ness. Positive effects of ants on plant fitness were less

commonly addressed than ant effects on herbivory, the

former representing only 21.5% of the independent com-

parisons included here. Only 18% of all studies reviewed

addressed effects of ant removal on herbivory and on

components of plant fitness simultaneously (Del-Claro

et al. 1996; Letourneau 1998; Izzo and Vasconcelos 2002;

Rudgers 2004). Among these, 62% were performed in

temperate systems. Effects of ant removal on plant repro-

duction were significant, but weaker than effects of ants on

plant herbivory. Plants without ants experienced a 25%

decrease in reproduction (flower, fruit, and seed production

combined: E??= -24.2, CI = -0.30 to –0.18, n = 58) and

effects were significant for tropical systems only. Weaker

effects of ants on plant reproduction might be explained

by several reasons: herbivores might not affect plant

fitness, as plants are able to tolerate herbivory without

fitness reduction (Strauss and Agrawal 1999). Alterna-

tively, effects of increased herbivory on ant-excluded

plants might not incur in fitness reductions in the same

season, as time lags for the beneficial effects of ants on

herbivory deterrence and subsequent results on plant

fitness might occur. The majority of studies reviewed

here were short-term studies (but see Torres-Hernandez

et al. 2000; Rudgers 2004) where aspects of plant

reproduction were evaluated soon after ant exclusion.

Long-term studies might be necessary to address the

relationship between ant presence, plant herbivory, and

plant reproduction.

Protective ant–plant or ant–hemipteran interactions are

important in both temperate and tropical communities

(Bronstein 1998; Heil and McKey 2003). Ant–plant mu-

tualisms in the neotropics have received much more

attention compared to other regions (Fiala et al. 1999) even

though predatory ants are considered keystone species in

temperate and boreal woodlands due to their effects on

herbivore community composition and abundance (Sipura

2002). In a recent review, Styrsky and Eubanks (2007) did

not observe significant differences in the effects of ant–

hemipteran interactions between temperate and tropical

regions. On the other hand, this study reviewed 81 studies

of the effects of ants as plant biotic defenses, and we have

shown stronger effects of ant removal on herbivory in

tropical environments compared to temperate ones. Pre-

sumably, this effect is due to higher ant and herbivore

diversity in the tropics as well as higher herbivore speci-

ficity (Coley and Barone 1996; Dyer et al. 2007; but see

Novotny et al. 2006) and ant aggressive behavior in trop-

ical compared to temperate regions. Moreover, tropical

areas harbor both myrmecophytes and myrmecophilies,

whereas temperate vegetation lacks myrmecophytic spe-

cies. In a review of tritrophic interactions in tropical and

temperate ecosystems, Dyer and Coley (2002) observed

that tropical plants are better defended against herbivores

then temperate species and that natural enemies have

strong negative effects on herbivores at all latitudes, but the

magnitude of the effect was significantly higher in tropical

than in temperate areas. Therefore, trophic cascades differ

between tropical and temperate ecosystems, and stronger

top-down effects of predators on herbivores and of herbi-

vores on plants, are significantly stronger in the tropics.

Our results have also shown that single ants on myr-

mecophytic plants have stronger effects on herbivory rates

and herbivore abundance than several ants, reinforcing the

idea that ants better protect tropical than temperate plants.

Our results have also shown that effects of single ant

species on arthropod communities and/or herbivore dam-

age and plant fitness were stronger than the effects of

several ant species. Visitation by multiple ant species that

vary in anti-herbivore abilities may result in reduced ben-

efits, relative to an exclusive association with a high-

quality mutualist (Miller 2007). The higher effect of single

ants in tropical plants could be explained by a stronger and

specialized mutualism, or by the presence of ecologically

dominant aggressive ants. Mosaics of behaviorally domi-

nant ants, for example, have been observed in tropical

canopies, where ants represent more than 90% of the

individuals and 50% of the arthropods (Dejean and Corbara

2003). Other studies also observed that plant resources,

such as EFNs, hemipterans, or domatia, shaping these ant

mosaics (Davidson 1997; Bluthgen et al. 2000; Hossaert-

McKey et al. 2001) influence thereby the outcomes of

546 Oecologia (2009) 160:537–549

123

plant–ant mutualisms in tropical regions. Defense of

exclusive foraging territories in tropical tree canopies by

particularly abundant and aggressive hemipteran-tending

ants reduces the density and diversity of other ants,

resulting in mosaic distributions of dominant and sub-

dominant arboreal ant species (Bluthgen et al. 2000, 2004;

Djieto-Lordon et al. 2007). Single but aggressive ants, such

as Azteca (Schultz and McGlynn 2000), have the potential

to influence the strength of the interaction between ants and

plants in mutualistic associations. However, the division

between dominant and subordinate ant species was not

possible using data provided by authors of the studies

reviewed. Also, the effect for single species was stronger

when ants from obligate association were removed, rein-

forcing the first hypothesis. The fact that almost 80% of the

myrmecophytic plants reviewed here were colonized by a

single ant species and myrmecophytic plants associated

with several ant species were rare in our review (only three

studies: see Alvarez et al. 2001; Heil et al. 2001; Michel-

angeli 2003) also reinforces the hypothesis of the strong

role of ants as plant biotic defenses in obligate ant–plants in

the tropical regions. In temperate regions, on the other

hand, all ant–plant associations are facultative and half of

these associations were maintained with single ant species

and half with several ants. Therefore, it is not surprising

that specialist ants have stronger effects than opportunistic

species.

The outcome of ant–plant mutualisms can be dependent

upon several biotic and abiotic factors (Heil and McKey

2003), and the type of food rewards are among those fac-

tors that shape this interaction. In this study, domatia-

bearing plants hosting honeydew-producing hemipterans

exhibited strongest effects of ant removal compared to

other ant attractors. In myrmecophiles, ants generally

exhibit low fidelity to the food association (Kersch and

Fonseca 2005) and many species are commonly present on

a given plant over its lifetime or might switch among plants

(Beattie 1985). In fact, EFN-based ant–plant associations

tend to be more generalized while the associations

involving domatia and food bodies tend to be more spe-

cialized and specific (Rico-Gray and Oliveira 2007).

Domatia seem to offer a much more specific reward than

EFNs (Heil and McKey 2003) and plant–ants generally

show strong fidelity to their ant–plants (Rico-Gray and

Oliveira 2007). Although widespread, mutualistic systems

involving either hemipterans or EFNs seem to be looser or

less specific with uncertain outcome to the plant (Rico-

Gray and Oliveira 2007) whereas aggressive ant behavior

on domatia-bearing plants has been previously recorded

(Janzen 1966; Heil and McKey 2003). Our results provide

support for the hypothesis that identity of the ant attractor

plays an important role in the evolution of the interaction

among plants, ants, and other herbivores, shaping therefore

the outcome of mutualistic interactions, especially in

tropical regions, where plants with domatia are commonly

found (Heil and McKey 2003).

Previous studies provided compelling evidence that ant–

hemipteran and ant–plant interactions can act as ‘keystone

interactions’ that dramatically change the structure of

arthropod communities on plants. In the presence of hon-

eydew-producing hemipterans, EFNs, food bodies, or

domatia, ants alter the abundance and distribution of spe-

cialist and generalist predators and parasitoids, and

multiple species of herbivores in several feeding guilds,

resulting in changes to local species diversity (Styrsky and

Eubanks 2007). We showed here that ant presence on

plants, regardless of the type of ant attractor or geographic

location, broadly affects the local abundance and distri-

bution of predators and insect herbivores, affecting in turn

herbivory levels and plant fitness. Therefore, ant–plant or

ant–hemipteran interactions may represent ‘keystone

interactions’ in many communities. However, the effect of

the ant presence is stronger in tropical environments, and

despite of the fact that selection has only rarely favored

obligate mutualisms (Rico-Gray and Oliveira 2007), the

positive outcome is also stronger in domatia-bearing plants

associated with single dominant ant species.

Acknowledgments This study was part of the course ‘‘Topics in

Ecology––Meta-Analysis’’ of the graduate program in Ecology,

Conservation and Management at UFMG, taught by T Cornelissen.

We would like to thank all the authors that kindly sent separates or

pdfs, especially Dr. Daniel Janzen. T Cornelissen acknowledges

FAPESP (06/57881-5) for a postdoctoral fellowship and G.W.

Fernandes acknowledges CNPq for a research fellowship (30.9633/

2007-9).

References

Alvarez G, Armbrecht I, Jimenez E, Armbrecht H, Ulloa-Chacon P

(2001) Ant–plant association in two Tococa species from a

primary rain forest of Colombian Choco (Hymenoptera: Form-

icidae). Sociobiology 38:585–602

A.P.G. [= Angiosperm Phylogeny Group] II (2003) An update of

the Angiosperm Phylogeny Group classification for the orders

and families of flowering plants: APG II. Bot.J Linn Soc

141:399–436

Beattie AJ (1985) The evolutionary ecology of ant–plant mutualisms.

Cambridge University Press, Cambridge

Bluthgen N, Verhaagh M, Goitıa W, Jaffe 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

Bluthgen N, Stork NE, Fiedler K (2004) Bottom-up control and co-

occurrence in complex communities: honeydew and nectar

determine a rainforest ant mosaic. Oikos 106:4344–4358

Bronstein JL (1998) The contribution of ant plant protection studies to

our understanding of mutualism. Biotropica 30:150–161

Bronstein JL, Alarcon R, Geber M (2006) The evolution of plant–

insect mutualisms. New Phytol 172:412–428

Oecologia (2009) 160:537–549 547

123

Bruna EM, Lapola DM, Vasconcelos HL (2004) Interspecific

variation in the defensive responses of obligate plant–ants:

experimental tests and consequences for herbivory. Oecologia

138:558–565

Coley PD, Barone JA (1996) Herbivory and plant defenses in tropical

forests. Annu Rev Ecol Syst 27:305–335

Compton SG, Robertson HG (1988) Complex interactions between

mutualisms: ants tending homopterans protect fig seeds and

pollinators. Ecology 69:1302–1305

Crutsinger GM, Sanders NJ (2005) Aphid-tending ants affect

secondary users in leaf shelters and rates of herbivory on Salixhookeriana in a coastal dune habitat. Am Midl Nat 154:296–304

Cuautle M, Rico-Gray V (2003) The effect of wasps and ants on the

reproductive success of the extrafloral nectaried plant Turneraulmifolia (Turneraceae). Funct Ecol 17:417–423

Davidson DW (1997) The role of resource imbalances in the

evolutionary ecology of tropical arboreal ants. Biol J Linnean

Soc 61:153–181

Davidson DW, McKey D (1993) The evolutionary ecology of

symbiotic ant–plant relationships. J Hymenoptera Res 2:13–83

Dejean A, Corbara B (2003) A review of mosaics of dominant ants in

rainforests and plantations. In: Basset Y, Novotny V, Miller SE

(eds) Arthropods of tropical forests: spatio-temporal dynamics

and resource use in the canopy. Cambridge University Press,

New York, pp 341–347

Dejean A, Delabie JHC, Cerdan P, Gibernau M, Corbara B (2006)

Are myrmecophytes always better protected against herbivores

than other plants? Biol J Linn Soc 89:91–98

Delabie JHC (2001) Trophobiosis between Formicidae and Hemiptera

(Sternorrhyncha and Auchenorrhyncha): an overview. Neotrop

Entomol 30:501–516

Delabie JHC, Osprina M, Zabala G (2003) Relaciones entre hormigas

y plantas: una introduccion. In: Fernandez F (ed) Introduccion a

las Hormigas de la region Neotropical. Instituto de Investigacion

de Recursos Biologicos Alexander Von Humboldt, Bogota, pp

167–180

Del-Claro K (2004) Multitrophic relationships, conditional mutual-

isms, and the study of interaction biodiversity in tropical

savannas. Neotrop Entomol 33:665–672

Del-Claro K, Oliveira PS (2000) Conditional outcomes in a neotrop-

ical treehopper–ant association: temporal and species-specific

variation in ant protection and homopteran fecundity. Oecologia

124:156–165

Del-Claro K, Berto V, Reu W (1996) Effect of herbivore deterrence

by ants on the fruit set of an extrafloral nectary plant, Qualeamultiflora (Vochysiaceae). J Trop Ecol 12:887–892

Del-Claro K, Byke J, Yugue GM, Morato MG (2006) Conservative

benefits in an ant–hemipteran association in the Brazilian

tropical savanna. Sociobiology 47:415–421

Djieto-Lordon C, Dejean A, Ring RA, Nkongmeneck BA, Lauga J,

McKey D (2007) Ecology of an improbable association: the

Pseudomyrmecine plant–ant Tetraponera tessmanni and the

myrmecophytic liana Vitex thyrsiflora (Lamiaceae) in Camer-

oon. Biotropica 37:421–430

Dutra HP, Freitas AVL, Oliveira PS (2006) Dual ant attraction in the

neotropical shrub Urera baccifera (Urticaceae): the role of ant

visitation to pearl bodies and fruits in herbivore deterrence and

leaf longevity. Funct Ecol 20:252–260

Dyer LA, Coley PD (2002) Tritrophic interactions in tropical and

temperate communities. In: Tscharntke T, Hawkins B (eds)

Multitrophic Level Interactions. Cambridge University Press,

Cambridge, pp 67–88

Dyer LA, Singer MS, Lill JT, Stireman JO, Gentry GL, Marquis RJ,

Greeney HF, Wagner DL, Morais HC, Diniz IR, Kursar TA,

Coley PD (2007) Host specificity of Lepidoptera in tropical and

temperate forests. Nature 448:696–699

Fiala B, Maschwitz U, Pong TY, Helbig AJ (1989) Studies of a

Southeast asian ant–plant association: protection of Macarangatrees by Crematogaster borneensis. Oecologia 79:463–470

Fiala B, Jakob A, Maschwitz U (1999) Diversity, evolutionary

specialization and geographic distribution of a mutualistic ant–

plant complex: Macaranga and Crematogaster in South East

Asia. Biol J Linn Soc 66:305–331

Fonseca CR (1994) Herbivory and the long-lived leaves of an

Amazonian ant-tree. J Ecol 82:833–842

Fowler SV, 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

Fraser AM, Axen AH, Pierce NE (2001) Assessing the quality of

different ant species as partners of a myrmecophilous butterfly.

Oecologia 129:452–460

Frederickson ME, Gordon DM (2007) The devil to pay: a cost of

mutualism with Myrmelachista schumanni ants in ‘devil’s

gardens’ is increased herbivory on Duroia hirsuta trees. Proc

R Soc Lond B 274:1117–1123

Freitas L, Galetto L, Bernardello G, Paoli AAS (2000) Ant exclusion

and reproduction of Croton sarcopetalus (Euphorbiaceae). Flora

195:398–402

Gaume L, Zacharias M, Grosbois V, Borges RM (2005) The fitness

consequences of bearing domatia and having the right ant

partner: experiments with protective and non-protective ants in a

semi-myrmecophyte. Oecologia 145:76–86

Gibb H (2003) Dominant meat ants affect only their specialist

predator in a complex natural system. Oecologia 136:609–615

Halaj J, Wise DH (2001) Terrestrial trophic cascades: how much do

they trickle? Am Nat 157:262–281

Heil M (2008) Indirect defense via tritrophic interactions. New Phytol

178:41–61

Heil M, McKey D (2003) Protective ant–plant interactions as model

systems in ecological and evolutionary research. Annu Rev Ecol

Evol Syst 34:425–453

Heil M, Fiala B, Maschwitz U, Linsenmair KE (2001) On benefits of

indirect defence: short- and long-term studies of antiherbivore

protection via mutualistic ants. Oecologia 126:395–403

Hossaert-McKey M, Orivel J, Labeyrie E, Pascal L, Delabie JHC,

Dejean A (2001) Differential associations with ants of three co-

occurring extrafloral nectary-bearing plants. Ecoscience 8:325–

335

Izzo TJ, Vasconcelos HL (2002) Cheating the cheater: domatia loss

minimizes the effects of ant castration in an Amazonian ant–

plant. Oecologia 133:200–205

James DG, Stevens MM, O’Malley KJ, Faulder RJ (1999) Ant

foraging reduces the abundance of beneficial and incidental

arthropods in Citrus canopies. Biol Control 14:121–126

Janzen DH (1966) Coevolution of mutualism between ants and

acacias in Central America. Evolution 20:249–275

Janzen DH (1967) Interaction of the bull’s horn Acacia (Acaciacornigera L.) with an ant inhabitant (Pseudomyrmex ferrugineaF. Smith) in Eastern Mexico. Univ Kans Sci Bull 47:315–558

Karhu KJ (1998) Effects of ant exclusion during outbreaks of a

defoliator and a sap-sucker on birch. Ecol Entomol 23:185–194

Kersch MF, Fonseca CR (2005) Abiotic factors and the conditional

outcome of an ant–plant mutualism. Ecology 86:2117–2126

Koptur S (2005) Nectar as fuel for plant protectors. In: Wackers FL,

van Rijn PCJ, Bruin J (eds) Plant-provided food for carnivorous

insects: a protective mutualism and its applications, Cambridge

University Press, Cambridge, pp 75–108

Letourneau DK (1998) Ants, stem-borers, and fungal pathogens:

Experimental tests of a fitness advantage in Piper ant–plants.

Ecology 79:593–603

Letourneau DK, Barbosa P (1999) Ants, stem borers, and pubescence

in Endospermum in Papua New Guinea. Biotropica 31:295–302

548 Oecologia (2009) 160:537–549

123

Messina FJ (1981) Plant protection as a consequence of an ant–

membracid mutualism: interactions on goldenrod (Solidago sp.).

Ecology 62:1433–1440

Michelangeli FA (2003) Ant protection against herbivory in three

species of Tococa (Melastomataceae) occupying different envi-

ronments. Biotropica 35:181–188

Miller TEX (2007) Does having multiple partners weaken the benefits

of facultative mutualism? A test with cacti and cactus-tending

ants. Oikos 116:500–512

Moller AP, Jennions MD (2001) Testing and adjusting for publication

bias. Trends Ecol Evol 16:580–586

Mooney KA (2007) Tritrophic effects of birds and ants on a canopy

food web, tree growth, and phytochemistry. Ecology 88:2005–

2014

Novotny V, Drozd P, Miller SE, Kulfan M, Janda M, Basset Y,

Weiblen GD (2006) Why are there so many species of

herbivorous insects in tropical rainforests? Science 313:1115–

1118

Offenberg J, Nielsen MG, Macintosh DJ, Havanon S, Aksornkoae S

(2005) Lack of ant attendance may induce compensatory plant

growth. Oikos 111:170–178

Oliveira PS (1997) The ecological function of extrafloral nectaries:

Herbivore deterrence by visiting ants and reproductive output in

Caryocar brasiliense (Caryocaraceae). Funct Ecol 11:323–330

Oliveira PS, Freitas AVL (2004) Ant–plant–herbivore interactions in

the Neotropical Cerrado Savanna. Naturwissenschaften 91:557–

570

Renault CK, Buffa LM, Delfino MA (2005) An aphid–ant interaction:

effects on different trophic levels. Ecol Res 20:71–74

Rico-Gray V, Oliveira P (2007) The ecology and evolution of ant–

plant interactions. University of Chicago Press, Chicago

Rosenberg MS, Adams DC, Gurevitch J (2000) MetaWin: statistical

software for meta-analysis. Version 2.0, Sinauer Associates,

Sunderland

Rosenthal R (1979) The ‘file drawer’ problem and tolerance for null

results. Psychol Bull 86:638–641

Rudgers JA (2004) Enemies of herbivores can shape plant traits:

Selection in a facultative ant–plant mutualism. Ecology 85:192–

205

Ruhren S (2003) Seed predators are undeterred by nectar-feeding ants

on Chamaecrista nictitans (Caesalpineaceae). Plant Ecol

166:189–198

Schemske DW (1982) Ecological correlates of a neotropical mutu-

alism: ant assemblages at Costus extrafloral nectaries. Ecology

63:932–941

Schmitz OJ, Hamback PA, Beckerman AP (2000) Trophic cascades in

terrestrial systems: a review of the effects of carnivore removal

on plants. Am Nat 55:141–153

Schultz TR, McGlynn TP (2000) The interaction of ants with another

organisms. In: Agosti D, Majer JD, Alonso LE, Schultz TR (eds)

Ants: standard methods for measuring and monitoring biodiver-

sity. Smithsonian Institution Press, Washington, pp 1–8

Sipura M (2002) Contrasting effects of ants on the herbivory and

growth of two willow species. Ecology 83:2680–2690

Strauss SY, Agrawal AA (1999) The ecology and evolution of plant

tolerance to herbivory. Trends Ecol Evol 14:179–185

Styrsky JD, Eubanks MD (2007) Ecological consequences of

interactions between ants and honeydew-producing insects. Proc

R Soc Lond Ser B Biol Sci 274:151–164

Suzuki N, Ogura K, Katayama N (2004) Efficiency of herbivore

exclusion by ants attracted to aphids on the vetch Viciaangustifolia L. (Leguminosae). Ecol Res 19:275–282

Torres-Hernandez L, Rico-Gray V, Castillo-Guevara C, Vergara JA

(2000) Effect of nectar-foraging ants and wasps on the repro-

ductive fitnesse ot Turnera ulmifolia (Turneraceae) in a coastal

sand dune in Mexico. Acta Zool Mex 81:13–21

Webber BL, Moog J, Curtis ASO, Woodrow IE (2007) The diversity

of ant–plant interactions in the rainforest understory tree,

Ryparosa (Achariaceae): food bodies, domatia, prostomata, and

hemipteran trophobionts. Biol J Linn Soc 154:353–371

Oecologia (2009) 160:537–549 549

123