Behav Ecol Sociobiol (1999) 45: 47±54 Ó Springer-Verlag 1999

ORIGINAL ARTICLE

Anurag A. Agrawal á Benjamin J. Dubin-Thaler

Induced responses to herbivory in the Neotropical ant-plant association

between Azteca ants and Cecropia trees: response of ants to potential
inducing cues

Received: 23 March 1998 / Accepted after revision: 5 July 1998

Abstract Plant defense against herbivores often involves responses to herbivory in obligate ant-plant systems, and
constitutive and inducible mechanisms of resistance. in facultative ant-plant associations, which may be more
Obligate ant-plants, which provide food and housing for common.
ants, are thought to primarily rely on ants for defense
against herbivores. This form of plant defense has Key words Ant-plant interactions á Induced
largely been viewed as static. We have been investigating resistance á Mutualism á Myrmecophyte á Plant cues
the dynamic nature of Azteca ants as an inducible de-
fense of Cecropia trees. Ants rapidly recruit to and pa-
trol sites of foliar damage. We propose that Azteca ants
can be viewed as an inducible defense for Cecropia trees Introduction
because of their sensitivity to cues associated with
herbivory, their rapid and aggressive recruiting ability, Ant-plants, plants that provide food and/or housing for
and their reclaimable and redeployable nature as a plant ants, are conspicuous members of tropical plant com-
defense. In this study, we examine ant behavior follow- munities, in part because their putative plant defenses
ing plant damage, and the potential cues that indude ant are visible on the outside of the leaves as opposed to
recruitment. We found that ants present on leaves when being hidden inside. Many ant-removal studies have
the plant is damaged leave the damaged leaf and recruit demonstrated that ants protect plants against herbivory
other ants to it, presumably by laying recruitment trails. and competing plants (Janzen 1966, 1969; Schupp 1986;
Volatile leaf cues associated with herbivory were im- Vasconcelos 1991; Fisher 1992; Fonseca 1994; Del-Claro
portant in eliciting an induced response in two experi- et al. 1996; Gaume et al. 1997; Oliveira 1997), and have
ments. However, we found that cues associated with a led to the hypothesis that ants are functionally equiva-
congeneric plant elicited a much stronger ant response lent to chemical defenses (Janzen 1966, 1969; Rehr et al.
than conspeci®c cues. Although the type of leaf damage 1973; McKey 1984, 1988; Davidson and Fisher 1991).
(gaping wounds versus leaf edge wounds) did not a€ect Ants are similar to chemical defenses in that their
the level of ant recruitment, the extent of damage did. numbers (concentrations) are dependent on various en-
Leaves with one hole punched showed a 50% increase in vironmental conditions (Agrawal and Rutter, in press).
ants, while leaves with ®ve holes punched in them elic- In addition, several authors have suggested that plant
ited a 100% increase in ant numbers. In sum, it appears defense by ants is not as simple as the presence/absence
that multiple plant-related cues associated with herbi- of ants, and that the dynamic behaviors of ants may be
vory are involved in induction of ant recruitment in the an important factor in their e€ectiveness (Letourneau
Cecropia-Azteca system. We discuss the generality of ant 1983; Madden and Young 1992; Rocha and Bergallo
1992; Gaume et al. 1997, Agrawal, in press; Agrawal and
Rutter, in press).
A.A. Agrawal (&) Some plant defenses are induced by herbivory. In-
e-mail: aaagrawal@ucdavis.edu, Tel.: +1-530-7527525 duced phytochemical and physical responses following
Fax: +1-530-7521537 herbivory have been well documented in temperate
Department of Entomology and Center for Population Biology
University of California at Davis, One Shields Avenue plants (Tallamy and Raupp 1991; Karban and Baldwin
Davis, CA 95616-8584, USA 1997). Few studies have investigated plant responses to
B.J. Dubin-Thaler
herbivory in tropical plants. Young (1987) found that
Department of Physics, Pupin Hall, Columbia University gira€e herbivory resulted in increased length of newly
New York, NY 10027, USA formed thorns on African Acacia drepanolobium.
48

Milewski et al. (1991) further demonstrated that this is provided in McDade et al. (1994). We conducted the ®eld work in
increase in thorn length resulted in reduced subsequent December 1997±January 1998 during the beginning of the dry
season.
grazing. Ants have also been studied as an inducible Of the early successional Neotropical ¯ora, the genus Cecropia
plant defense (Fiala and Maschwitz 1990; Madden and (Urticaceae; Judd et al. 1994) contains over 100 species, most of
Young 1992; Rocha and Bergallo 1992; Cronin 1998; which have associations with symbiotic ants. Azteca spp. (For-
Agrawal, in press). Belt (1874) was one of the ®rst nat- micidae: Dolichoderinae) ants inhabit most of the C. obtusifolia
trees at La Selva Biological Station. The mutualism between Ce-
uralists to document that disturbance to myrmecophytic cropia spp. and Azteca spp. has been intensively studied (Skutch
plants resulted in aggressive biting, stinging, and 1945; Janzen 1969, 1973; Schupp 1986; Jolivet 1988, 1990; Longino
swarming of the ants living inside ant-plants. The gen- 1989, 1991; Davidson and Fisher 1991; Rocha et al. 1992; Folgarait
eral biology of ants makes them well suited for use as an and Davidson 1995; Vasconcelos and Casimiro 1997; Yu and
inducible plant defense: (1) they have acute sensory Davidson 1997). Like almost all members of the genus, C. obtusi-
folia develops hollow stems in which ant colonies nest. C. obtusi-
mechanisms to detect disturbance and chemical cues, (2) folia also produces glycogen-containing MuÈllerian food bodies and
they are often aggressive and have well-developed de- pearl bodies, which may contain carbohydrates, lipids, and proteins
fense mechanisms (e.g., biting and stinging), (3) they (Rickson 1971, 1973; O'Dowd 1982; A.A. Agrawal, personal ob-
have recruiting mechanisms which make them rapidly servation). Azteca ants consume the food bodies, patrol the plant
and, in many cases, successfully reduce herbivory and increase the
deployable and also reclaimable, and (4) individual ants ®tness of plants relative to conspeci®cs without ant protectors
will give their life in defense of the colony or its resources (Schupp 1986; Rocha and Bergallo 1992; Vasconcelos and Casi-
(HoÈlldobler and Wilson 1990). miro 1997). Schupp (1986) demonstrated that when its ant inhab-
We have been studying the aggressive behavior of itants were removed, C. obtusifolia experienced higher herbivory,
increased vine cover, and signi®cantly stunted growth. We studied
Azteca spp. ants that inhabit Neotropical Cecropia ob- young C. obtusifolia trees inhabited by Azteca xanthocroa or Azteca
tusifolia trees. Leaf damage resulted in a tripling of the constructor. Both ant species are extremely aggressive obligate
number of ants patrolling damaged leaves compared to mutualists that defend Cecropia spp. and are dicult to distinguish
undamaged controls (Agrawal, in press), and ant num- from each other without examining the queen (J. Longino, personal
ber is positively associated with resistance against her- communication). Worker ants were sampled from each tree to
make sure that the ants were not Azteca alfari, the apparently less
bivores in Cecropia trees (Rocha and Bergallo 1992). In aggressive and less defensive species of Azteca ant that inhabits
this study, we extend our previous results and examine Cecropia spp. (Longino 1991; but see Vasconcelos and Casimiro
several potential cues which may serve as signals that 1997). A. alfari can be easily distinguished by examining workers (J.
ants use to recruit to damaged leaves. Our goal is to Longino, personal communication).
understand how ants respond to leaf damage, and
whether their response is largely due to signals produced
by the plant, or to other cues associated with herbivory Recruiting behavior
such as disturbance and leaf damage.
Our previous studies of induction in the Cecropia-Azteca system
Our previous work suggested that volatiles associated demonstrated that leaf damage caused a ®vefold increase in the
with C. obtusifolia may be important in the induction number of Azteca spp. ants on damaged leaves of C. obtusifolia
process and that the type of damage may also in¯uence compared to that on disturbed but undamaged control leaves on
the level and persistence of ant patrolling. In this study the same tree (Agrawal, in press). Although we initially attributed
this di€erential accumulation of ants on the damaged leaf to in-
we asked: what is the behavioral mechanism of ant in- duction of ant ``recruitment,'' we did not have data on the mech-
duction on damaged leaves? It is important to distin- anism of induction. Wilson (1971) de®ned recruitment as
guish between a plant-wide ant response to herbivory communication between nestmates that results in movement of ants
and active ``recruitment'' of ants at the damage site. Our towards an area where work is required. Recruitment can be trig-
gered by a variety of mechanisms, of which chemical trail com-
study is the ®rst to speci®cally test for ant recruitment munication is the most prevalent (HoÈlldobler and Wilson 1990). As
per se to damaged leaves. We then conducted experi- an alternative to recruitment per se, ants may respond systemically
ments to determine the e€ects of (1) plant chemicals to the induction cues and simply accumulate on damaged leaves.
speci®c to C. obtusifolia, (2) the type of leaf damage, and To determine if the ants were indeed recruiting to sites of foliar
(3) the extent of leaf damage. damage, we conducted experiments where we in¯icted leaf damage
(®ve holes from a standard hole puncher, approximately 30 mm2
each) to individual leaves of plants and intensively observed ant
activity. We recorded every ant that came onto or left the leaf at the
leaf-petiole juncture. Similarly, we recorded every ant that came
onto or left the petiole at the petiole-stem juncture (see Fig. 1).
Methods These measurements were taken simultaneously for 2 min prior to
the leaf damage and for 12 min following leaf damage. We re-
corded ant numbers at the end of every 30-s period during this time
Study site and natural history so that we could construct a 30-s ¯ux index for the net change in the
number of ants. We conducted these observations on ®ve trees and
This study was conducted at the Organization for Tropical Studies report the results of a representative tree since the results were
La Selva Biological Station, near the town of Puerto Viejo de consistent among trees. We performed a paired t-test on the mean
SarapiquõÂ , Heredia Province, Costa Rica (10°26¢N and 84°00¢W). number of pre-damage ants and the total number of ants 12 min
La Selva is located in a lowland tropical wet forest on the Costa after damage. Twelve minutes post-damage was previously deter-
Rican Caribbean slope and receives approximately 4000 mm of mined to be the peak time of ant activity following damage
rainfall each year. Detailed information about the site and climate (Agrawal, in press).
 49

tempted to create two types of damage with an equal amount

of leaf tissue damage, but di€ering in the visual cues left
behind. On one leaf of each of 20 plants we punched 15 holes
(30 mm2 each), to create approximately 30 linear cm of
damaged leaf and leaving behind gaping holes. On the paired
leaf of each hole-punched leaf we damaged 30 linear cm of
leaf by removing (using scissors) a thin (less than 0.5-cm) strip
of leaf tissue following the outline of the leaf perimeter.
(5) To determine the importance of the extent of damage in ant
recruitment, we conducted experiments in which paired leaves
of ten trees received either one or ®ve hole punches.

Statistical analysis

All analyses were conducted using the MGLH routine in Systat

(Wilkinson et al. 1992). All experiments were conducted with a
paired design, where each tree or experimental unit had both treat-
ments, with no replication within each unit. Repeated-measures
analysis of variance procedures were used to test for treatment ef-
fects in all experiments. Treatments were considered ®xed e€ects and
tree was considered a random e€ect. In such designs, the between-
subjects (remainder) mean square and degrees of freedom is used in
the denominator to calculate the F-statistic (Zar 1996, p. 267).

Fig. 1 Schematic diagram of the experimental design used to test for

the behavioral responses of ants to leaf damage. A represents the Results
petiole-stem juncture and B represents the leaf-petiole juncture. Points
A and B were 30±60 cm apart
Recruiting behavior

Inducing signals In Fig. 2, we present the mean change in ant numbers

every 30 s at the leaf-petiole and the petiole-stem junc-
To test for the e€ects of various cues associated with herbivory
on inducing ant recruitment, we monitored ant numbers on
control and treatment C. obtusifolia leaves using a paired-leaf
design, with each tree having a treatment and a control leaf.
For each experiment we selected 10±20 young trees between 1.5
and 2.5 m tall, lacking signs of recent herbivory, and with es-
tablished colonies of A. xanthocroa or A. constructor. On each
tree, we chose two of the youngest, fully expanded leaves
coming o€ of the main stem (about 1±2 m apart) and randomly
assigned one to the ``treatment'' and one to the ``control''
group. We censused the number of ants on the leaves prior to
imposing the treatment and at 4, 8, 12, and 16 min after
damage. The speci®c manipulations follow:

(1) To test for e€ects of leaf volatiles on ant recruitment, we

augmented the treatment leaf with a freshly cut C. obtusifolia
leaf disc (9 cm2, cut from a di€erent tree), and the control leaf
with a disc of paper cut to the same size …n ˆ 13 pairs).
(2) To test for the e€ect of conspeci®c versus heterospeci®c leaf
volatiles, we augmented one leaf with a freshly cut C. obtusi-
folia leaf disc (9 cm2, cut from a di€erent tree) and the other
leaf with a leaf disc cut from a another myrmecophytic Ce-
cropia species common at La Selva Biological Station, C. in-
signis (n ˆ 19 pairs). At La Selva, C. insignis (like C.
obtusifolia) is often inhabited by A. xanthocroa and A. con-
structor.
(3) As an additional test of the importance of leaf volatiles, we
tested for the e€ects of a methanol extract of C. obtusifolia on
ant recruitment. We extracted 15 g of fresh C. obtusifolia foliage
by cutting the foliage into approximately 1-cm2 pieces and
soaking them in 100 ml of methanol for 24 h; after 24 h, the leaf
pieces were removed. Methanol generally extracts both polar
and non-polar components of the foliage. On each of 19 trees
we added ®ve drops of the C. obtusifolia extract to the treatment Fig. 2a±b The net change in numbers of ants at the leaf-petiole (A)
leaf and ®ve drops of methanol onto the control leaf. and the petiole-stem (B) juncture every 30 s (solid line). Five hole
(4) Previous results suggested that the type of leaf damage may punches were imposed on the leaf at time 0. The dashed line represents
be important in the strength and/or persistence of ant re- the total (cumulative) number of ants that were on the leaf at any
cruitment (Agrawal, in press). In this experiment, we at- given time
50

ture for a representative damaged plant. The observed pelled the ants, and that the C. obtusifolia extract tem-
pattern indicates a net decrease in the number of ants at pered that response (Fig. 4, Table 1). Volatiles from
each juncture prior to each major in¯ux of ants. For conspeci®c plants of C. obtusifolia resulted in approxi-
example, prior to the ®rst major in¯ux of ants to the mately 30% more ants than on control leaves in both
damaged leaves, in every sampled tree we observed two experiments. Heterospeci®c volatiles from C. insignis
to ®ve ants leave the leaf-petiole juncture and also leave induced much greater ant recruitment (165%) than
the petiole-stem juncture. In all cases, this out¯ow of volatiles from C. obtusifolia (Fig. 3B, Table 1).
ants was almost immediately followed by a large net The type of damage (gaping wounds vs leaf edge
in¯ow onto the leaves. We interpret this behavior as wounds) did not in¯uence ant response to damage, and
``recruitment;'' it appeared that ants needed to leave the in both treatments, ant numbers approximately tripled
leaves and travel all the way down the petiole o€ to the following damage (Fig. 5A, Table 1). Extent of damage
stem before there was a in¯ux of ants to the leaf. This had a signi®cant e€ect on the level of Azteca ant re-
experiment also con®rmed our previous results indicat- cruitment (Figure 5b, Table 1). On leaves with one hole
ing a net increase in the number of ants 12 min after punch ant numbers increased approximately 50%, while
damage (mean number ‹ SE of ants per leaf: pre- on leaves with ®ve hole punches, ant numbers increased
damage ˆ 10.0 ‹ 2.1, post-damage ˆ 34.0 ‹ 5.3; t ˆ just over 100%.
6.303, df ˆ 4, P ˆ 0.003).

Inducing signals Discussion

Azteca ants responded to plant volatiles and responded Induced responses to herbivory are common in most
di€erently to conspeci®c and heterospeci®c volatiles. groups of plants including annual and perennial species
Volatiles from C. obtusifolia (leaf disc experiment) and both terrestrial and aquatic species (Karban and
caused signi®cant induction of ant recruitment relative Baldwin 1997). Plant responses to herbivory can take the
to paper controls (Fig. 3A, Table 1). In the methanol form of phytochemical changes such as increases in
extract experiment, it appears that methanol alone re- proteinase inhibitors, oxidative enzymes, and alkaloids
following wounding in solanaceous plants (Green and
Ryan 1972; Baldwin 1991; Du€ey and Stout 1996; Stout
et al. 1996). Plant responses can also take the form of
increased density or size of physical resistance mecha-
nisms such as thorns, spines, and trichomes (Young
1987; Baur et al. 1991; Myers and Bazely 1991;Young
and Okello 1998; A.A. Agrawal, unpublished data).
Induced responses have also been reported in ant-plants,
most often as changes in extra¯oral nectar volume or
composition following herbivory (Mound 1962; Ste-
phenson 1982; Koptur 1989; Smith et al. 1990; Agrawal
and Rutter, in press). These examples of nectar re-
sponses following herbivory were all found in non-ob-
ligate ant-plant species, and ant responses were not
quanti®ed in these studies. In more obligate systems,
there have been observations of ant responses to dis-
turbance and herbivory, although the mechanisms are
unknown in each case (Crematogaster-Macaranga: Fiala
and Maschwitz 1990; Camponotus-Macaranga: Federle
et al. 1998; Crematogaster-Acacia: Madden and Young
1992; Azteca-Cecropia: Rocha and Bergallo 1992;
Agrawal, in press; Pseudomyrmex-Acacia: Cronin 1998).
For example, Fiala and Maschwitz (1990) found that
ants on Asian Macaranga trees showed a strong nu-
merical response to leaf damage and increased in num-
bers for up to 3 h following damage. Similarly, we have
shown in our previous work with Cecropia trees that leaf
damage per se was an important cue for induction of ant
recruitment, and that disturbance alone caused minimal
Fig. 3a±b Recruitment of ants to paired leaves that were treated with
a paper disc versus a leaf disc of Cecropia obtusifolia (A), or a leaf disc
induction (Agrawal, in press). Induction of ants in these
of C. obtusifolia versus a leaf disc of C. insignis (B). Data shown are obligate systems does not appear to be associated with
the mean + SE induction of ant rewards such as nectar or food bodies.
 51

Table 1 Repeated-measures analysis of variance tables for the e€ects of various stimuli on recruitment of Azteca ants on Cecropia
obtusifolia trees. Probabilities corrected for sphericity are provided using the Greenhouse-Geisser correction (G-G)

Experiment Source df MS F P G-G

Leaf disc vs paper

Treatment 1 238.784 4.418 0.05
Tree 18 747.947 13.838 <0.001
Error 18 54.051
Time 4 72.039 9.718 <0.001 <0.001
Time ´ treatment 4 19.771 2.667 0.039 0.071
Time ´ tree 72 21.787 2.939 <0.001 <0.001
Error 72 7.413
C. obtusifolia vs C. insignis
Treatment 1 1794.531 5.531 0.037
Tree 12 3232.869 9.964 <0.001
Error 12 324.464
Time 4 582.873 17.469 <0.001 <0.001
Time ´ treatment 4 97.204 2.913 0.031 0.061
Time ´ tree 48 81.294 2.436 0.001 0.009
Error 48 33.366
Extracts
Treatment 1 188.005 5.409 0.032
Tree 18 1541.714 44.352 <0.001
Error 18 34.761
Time 4 13.592 1.103 0.362 0.346
Time ´ treatment 4 20.913 1.697 0.160 0.195
Time ´ tree 72 17.095 1.387 0.084 0.156
Error 72 12.321
Type of damage
Treatment 1 389.205 1.379 0.255
Tree 19 9778.792 34.655 <0.001
Error 19 282.173
Time 4 3029.683 45.061 <0.001 <0.001
Time ´ treatment 4 32.793 0.488 0.745 0.588
Time ´ tree 76 322.219 4.792 <0.001 <0.001
Error 76 67.235
Extent of damage
Treatment 1 784.000 13.972 0.005
Tree 9 1141.107 20.337 <0.001
Error 9 56.111
Time 4 563.840 44.031 <0.001 <0.001
Time ´ treatment 4 43.250 3.337 0.019 0.032
Time ´ tree 36 57.207 4.467 <0.001 <0.001
Error 36 12.806

It is generally thought that inducible defenses in- have costs, so that induction decreases plant ®tness in
crease plant ®tness in the presence of herbivores, but the absence of herbivores (Harvell 1990; Karban and
Baldwin 1997; Agrawal 1998a). Such a cost-bene®t
trade-o€, if genetically based, may constrain the evolu-
tion of inducible defenses. For ant-plant systems, espe-
cially those that are facultative, i.e., where plants o€er
an inducible reward to ants following herbivory (see
above), such cost-bene®t models may be appropriate.
However, in obligate systems where ants may ``induce
themselves,'' alternative hypotheses need to be consid-
ered (Agrawal and Rutter, in press). In many of these
obligate systems, the ants remove herbivores from
plants, but do not consume them. In such systems, it is
not clear who or what regulates the induction, the ants
or the plants. Ants do not appear to directly bene®t from
inducing recruitment and removing herbivores. How-
ever, if herbivory a€ects plant performance and ant ®t-
Fig. 4 Recruitment of ants to paired leaves that were treated with ®ve ness is tightly coupled to plant performance, then ants
drops of a methanol extract of C. obtusifolia versus methanol alone. may ``induce themselves'' in response to herbivory and
Data shown are the mean + SE may not be responding to plant cues (Agrawal and
52

leaves and used them to elicit ant recruitment on another

tree, we observed no di€erence in recruitment of ants to
leaves adjacent to damaged leaves compared to recruit-
ment on leaves adjacent to undamaged leaves (Agrawal,
in press).
Azteca ants responded to host plant volatiles. These
volatiles may serve as a cue that herbivores are present
on the plant. Fiala and Maschwitz (1990) also reported
ant responses to host plant cues in the Crematogaster-
Macaranga system. Interestingly, they observed the
greatest ant response to conspeci®c cues from water
extracts of the foliage, compared to extracts from con-
generic and unrelated plants. However, ants also re-
spond to plant cues in order to ``prune'' competing
plants and encroaching vines (Janzen 1969; Davidson et
al. 1988). Fiala and Maschwitz observed the strongest
``pruning'' response to non-host plant leaves. Our results
can be interpreted consistently with the observations of
Fiala and Maschwitz (1990): we found that ants re-
sponded to host plant volatile cues (extract, and leaf disc
experiments), but that they had a stronger response to
heterospeci®c cues in the leaf disc experiment. The
stronger response of ants to C. insignis may be due to the
ants' perception of this plant as a competitor which they
were attempting to prune. Alternatively, C. insignis,
which is an ant-plant which also has an association with
Fig. 5a±b Recruitment of ants to paired leaves that were treated with A. xanthocroa and A. constructor, may produce similar
hole punches versus scissors clipping around the perimeter of the leaf alarm signals that are stronger than those present in C.
(equal amount of linear length of leaf damage in both treatments) (A), obtusifolia. Further work on volatile leaf cues from
and one hole punch versus ®ve hole punches (B). Data shown are the various related and unrelated species will be instructive
mean + SE
in resolving how ants recognize host plant wounding
versus competitor encroachment. In addition, this e€ect
of the ants' apparent response to plant cues may also be
Rutter, in press). In such cases, the plant may not o€er related to the territoriality of arboreal ants (see David-
additional rewards to induce ant recruitment. Our in- son et al 1988; Adams 1994).
vestigations of the Cecropia-Azteca system have focused The current study indicates that visual or tactile cues
on how the ants respond to damage and what the in- are less important than chemical cues. In previous ex-
ducing cues may be. periments we had hypothesized that visual and/or tactile
Azteca ants appear to respond to foliar damage by cues were important. Treating a C. obtusifolia leaf with
alarming nestmates and show a numerical increase pri- several hundred pin wounds induced signi®cant ant re-
marily due to some form of recruitment. In our obser- cruitment beyond the recruitment observed on paired
vations of ant behavior, the ants present on the leaves at leaves that were disturbed but not damaged (Agrawal, in
the time of damage immediately became excited and press). However, the magnitude and persistence of this
began swarming around the damaged leaf and investi- recruitment was much weaker than that observed fol-
gating the damaged sites. Ant numbers on the leaf, lowing hole punch damage. This led us to the hypothesis
however, did not increase for a few minutes, and the that gaping holes associated with hole punching were
increase was always preceded by a fraction of the ants on important cues not present in the pin-wounding treat-
the leaf leaving and travelling all the way o€ the petiole ment. Our experiments with leaf discs and methanol
to the stem. These ants were often observed walking extracts resulted in ant responses to conspeci®c volatiles
with a curled gaster, perhaps indicating that the ants even though treatment and control leaves had the same
were releasing alarm substances or trail pheromones visual and tactile cues. Furthermore, ants showed simi-
(P.S. Ward, personal communication, see also Wheeler lar numerical responses to clipping of the leaf perimeter
et al. 1975). The alternative to the ``recruitment hy- and hole punching when the linear length of damage was
pothesis'' is that ants not present on the damaged leaf controlled for. Our previous results may have been
respond to plant cues that emanate from damaged caused by a di€erence in the actual area of leaf tissue
leaves. This is unlikely because volatile damage cues are that was damaged in the pin-wounding versus hole-
likely to di€use and not reach the rest of the colony. punching experiments.
Consistent with this, in a previous experiment in which Extent of damage appears to be an important deter-
we removed damaged and undamaged C. obtusifolia minant of the ant response to damage. Induced phyto-
 53

chemical responses also appear to be positively associ- Baldwin IT, Schmelz EA (1994) Constraints on an induced defense:
ated with the level of herbivory (Karban 1987; Lin et al. the role of leaf area. Oecologia 97:424±430
Baur R, Binder S, Benz G (1991) Nonglandular leaf trichomes as
1990, Bodnaryk 1992; Baldwin and Schmelz 1994). In short-term inducible defense of the gray alder, Alnus incana
our experiments, we found that the ant response to ®ve (L.), against the chrysomelid beetle, Agelastica alni L. Oecolo-
hole punches was twice that of the response to one hole gia 87:219±226
punch. This result has two implications. First, the Belt T (1874) The naturalist in Nicaragua. Dent, London
Bodnaryk RP (1992) E€ects of wounding on glucosinolates in the
amount of damage can have a quantitative e€ect on ant cotyledons of oilseed rape and mustard. Phytochemistry
recruitment. We observed ant responses to the single 31:2671±2677
hole punch (Fig. 5b), indicating that both our experi- Cronin G (1998) Between-species and temporal variation in acacia-
mental levels of damage fell above the threshold of ant-herbivore interactions. Biotropica 30:135±139
damage necessary to cause induction. Rocha and Ber- Davidson DW, Fisher BL (1991) Symbiosis of ants with Cecropia
as a function of light regime. In: Huxley CR, Cutler DF (eds)
gallo (1992) additionally found that the extent of Azteca Ant-plant interactions. Oxford University Press, Oxford, pp
responses to herbivores was positively correlated with 289±309
subsequent resistance to herbivores. The second impli- Davidson DW, Longino JT, Snelling RR (1988) Pruning of host
cation of our result is that ant responses do not appear plant neighbors by ants: an experimental approach. Ecology
69:801±808
to be simply additive with regard to the level of damage. Del-Claro K, Berto V, Reu W (1996) E€ect of herbivore deterrence
A ®vefold increase in damage only caused a twofold by ants on the fruit set of an extra¯oral nectary plant, Qualea
increase in ant recruitment. Detailed studies on the im- multi¯ora (Vochysiaceae). J Trop Ecol 12:887±892
portance of the extent of damage will be needed to Du€ey SS, Stout MJ (1996) Antinutritive and toxic components of
plant defense against insects. Arch Insect Biochem Physiol
characterize the density-dependent relationship between 32:3±37
herbivory, induced responses, and induced resistance to Federle W, Maschwitz U, Fiala B (1998) The two-partner ant-plant
herbivores. system of Camponotus (Colobopsis) sp 1 and Macaranga pun-
cticulata (Euphorbiaceae): natural history of the exceptional ant
partner. Insectes Soc 45:1±16
Fiala B, Maschwitz U (1990) Studies on the South East Asian ant-
Conclusion plant association Crematogaster borneensis/Macaranga: adap-
tations of the ant partner. Insectes Soc 37:212±231
Resident Azteca ants on C. obtusifolia trees respond to Fisher BL (1992) Facultative ant association bene®ts a Neotropical
volatile plant cues associated with herbivory. Such vol- orchid. J Trop Ecol 8:109±114
Folgarait PJ, Davidson DW (1995) Myrmecophytic Cecropia ±
atile cues may be released in a fashion that is density antiherbivore defenses under di€erent nutrient treatments. Oe-
dependent with the level of damage. Excited ants recruit cologia 104:189±206
nestmates to damage sites. Induction of ant recruitment Fonseca CR (1994) Herbivory and the long-lived leaves of an
in this system is likely to serve as a mechanism of in- Amazonian ant-tree. J Ecol 82:833±842
Gaume L, McKey D, Anstett MC (1997) Bene®ts conferred by
duced resistance to herbivores. ``timid'' ants: active anti-herbivore protection of the rainforest
tree Leonardoxa africana by the minute ant Petalomyrmex
Acknowledgements We thank Louis LaPierre for sharing his ®eld phylax. Oecologia 112:209±216
sites and experimental trees and Lee Dyer, Grant Gentry, and Green TR, Ryan CA (1972) Wound induced proteinase inhibitor in
Deborah Letourneau for sharing their expertise and ideas. Jennifer plant leaves: a possible defense mechanism against insects.
Thaler, Pat Dubin, and Ed Wurtz provided helpful discussions in Science 175:776±777
the ®eld. The manuscript was improved by the thoughtful com- Harvell CD (1990) The ecology and evolution of inducible de-
ments of Jennifer Thaler, Sean Brady, Rick Karban, Truman fenses. Q Rev Biol 65:323±340
Young, Lynn Adler, Lee Dyer, Phil Ward, Pekka Pamilo, and two HoÈlldobler B, Wilson EO (1990) The ants. Belknap Press, Cam-
anonymous reviewers. Many thanks to Lynn S. Adler for drawing bridge, Mass
Fig. 1. This study was funded by an Organization for Tropical Janzen DH (1966) Coevolution of mutualism between ants and
Studies Graduate Fellowship, a National Science Foundation acacias in Central America. Evolution 20:249±275
Doctoral Dissertation Improvement Grant (DEB-9701109), and Janzen DH (1969) Allelopathy by myrmecophytes: the ant Azteca
the Center for Population Biology at the University of California at as an allelopathic agent of Cecropia. Ecology 50:147±153
Davis. The experiments presented in this paper comply with the Janzen DH (1973) Dissolution of mutualism between Cecropia and
current laws of Costa Rica. its Azteca ants. Biotropica 5:15±28
Jolivet P (1988) A case of parasitism of an ant-Cecropia mutualism
in Brazil. Adv Myrmecol 1:327±335
Jolivet P (1990) Relative protection of Cecropia trees against leaf-
References cutting ants in tropical America. In: Vander Meer RK, Ja€e K,
Cedeno A (eds) Applied myrmecology: a world perspective.
Adams ES (1994) Territory defense by the ant Azteca trigona: Westview, Boulder, Colo, pp 251±254
maintenance of an arboreal ant mosaic. Oecologia 97:202±208 Judd WS, Sanders RW, Donoghue MJ (1994) Angiosperm family
Agrawal AA (in press) Leaf damage and associated cues induced pairs: preliminary phylogenetic analysis. Harvard Pap Bot 5:1±
aggressive ant recruitment in a Neotropical ant-plant. Ecology 53
Agrawal AA (1998a) Induced responses to herbivory and increased Karban R (1987) Environmental conditions a€ecting the strength of
plant performance. Science 279:1201±1202 induced resistance against mites in cotton. Oecologia 73:414±419
Agrawal AA, Rutter MT (in press) Dynamic anti-herbivore defense Karban R, Baldwin IT (1997) Induced responses to herbivory.
in ant-plants: the role of induced responses. Oikos University of Chicago Press, Chicago
Baldwin IT (1991) Damage-induced alkaloids in wild tobacco. In: Koptur S (1989) Is extra¯oral nectar production an inducible de-
Tallamy DW, Raupp MJ (eds) Phytochemical induction by fense? In: Bock JH, Linhart YB (eds) The evolutionary ecology
herbivores. Wiley, New York, pp 47±69 of plants. Westview, Boulder, Colo, pp 323±339
54

Letourneau DK (1983) Passive aggression: an alternative hypoth- Rocha CFD, Bergallo HG (1992) Bigger ant colonies reduce
esis for the Piper-Pheidole association. Oecologia 60:122±126 herbivory and herbivore residence time on leaves of an ant-
Lin H, Kogan M, Fischer D (1990) Induced resistance in soybean plant: Azteca muelleri vs. Coelomera ru®cornis on Cecropia
to the Mexican bean beetle (Coleoptera: Coccinellidae): com- pachystachya. Oecologia 91:249±252
parisons of inducing factors. Environ Entomol 19:1852±1857 Schupp EW (1986) Azteca protection of Cecropia: ant occupation
Longino JT (1989) Geographic variation and community structure bene®ts juvenile trees. Oecologia 70:319±385
in an ant-plant mutualism: Azteca and Cecropia in Costa Rica. Skutch AF (1945) The most hospitable tree. Sci Mon 60:5±17
Biotropica 21:126±132 Smith LL, Lanza J, Smith GC (1990) Amino acid concentrations in
Longino JT (1991) Azteca ants in Cecropia trees: taxonomy, colony extra¯oral nectar of Impatiens sultani increase after simulated
structure, and behaviour. In: Huxley CR, Cutler DF (eds) Ant- herbivory. Ecology 71:107±115
plant interactions. Oxford University Press, Oxford, pp 271± Stephenson AG (1982) The role of the extra¯oral nectaries of
288 Catalpa speciosa in limiting herbivory and increasing fruit
Madden D, Young TP (1992) Symbiotic ants as an alternative production. Ecology 63:663±669
defense against gira€e herbivory in spinescent Acacia drepano- Stout MJ, Workman KV, Du€ey SS (1996) Identity, spatial dis-
lobium. Oecologia 91:235±238 tribution, and variability of induced chemical responses in to-
McDade LA, Bawa KS, Hespenheide HA, Hartshorn GS (1994) La mato plants. Entomol Exp Appl 79:255±271
Selva. University of Chicago Press, Chicago Tallamy DW, Raupp MJ (1991) Phytochemical induction by her-
McKey D (1984) Interactions of the ant-plant Leonardoxa africana bivores. Wiley, New York
(Caesalpiniaceae) with its obligate inhabitants in a rainforest in Vasconcelos HL (1991) Mutualism between Maieta guianensis
Cameroon. Biotropica 16:81±99 Aubl., a myrmecophytic melastome, and one of its ant inhabi-
McKey D (1988) Promising new directions in the study of ant-plant tants: ant protection against insect herbivores. Oecologia
mutualisms. In: Greuter W, Zimmer B (eds) Proceedings of the 87:295±298
XIV International Botanical Congress. Koeltz, Konigstein/ Vasconcelos HL, Casimiro AB (1997) In¯uence of Azteca alfari
Taunus, pp 335±355 ants on the exploitation of Cecropia trees by a leaf-cutting ant.
Milewski, AV, Young TP, Madden D (1991) Thorns as induced Biotropica 29:84±92
defenses: experimental evidence. Oecologia 86:70±75 Wheeler JW, Evans SL, Blum MS, Torgerson RS (1975) Cyclo-
Mound LA (1962) Extra-¯oral nectaries of cotton and their se- pentyl ketones: identi®cation and function in Azteca ants. Sci-
cretions. Empire Cotton Growing Rev 39:254±260 ence 187:254±255
Myers JH, Bazely D (1991) Thorns, spines, prickles, and hairs: are Wilkinson L, Hill M, Welna JP, Birkenbeuel GK (1992) SYSTAT
they stimulated by herbivory and do they deter herbivores. In: for Windows: statistics, version 5 Edition. SYSTAT Inc, Ev-
Tallamy DW, Raupp MJ (eds) Phytochemical induction by anston, Ill
herbivores. Wiley, New York, pp 325±344 Wilson EO (1971) The insect societies. Belknap, Cambridge, Mass
O'Dowd DJ (1982) Pearl bodies as ant food: an ecological role for Young TP (1987). Increased thorn length in Acacia drepanolo-
some leaf emergences of tropical plants. Biotropica 14:40±49 bium ± an induced response to browsing. Oecologia 71:436±
Oliveira PS (1997) The ecological function of extra¯oral nectaries: 438
herbivore deterrence by visiting ants and reproductive output Young TP, Okello BD (1998). Relaxation of an induced defense
in Caryocar brasiliense (Caryocaraceae). Funct Ecol 11:323± after exclusion of herbivores. Oecologia 115:508±513
330 Yu DW, Davidson DW (1997) Experimental studies of species-
Rehr SS, Feeny PP, Janzen DH (1973) Chemical defence in Central speci®city in Cecropia-ant relationships. Ecol Monogr 67:273±
American non-ant-acacias. J Anim Ecol 42:405±416 294
Rickson FR (1971) Glycogen plastids in MuÈllerian body cells of Zar JH (1996) Biostatistical analysis, 3rd edn. Prentice Hall, Upper
Cercopia peltata ± a higher green-plant. Science 173:344±347 Saddle River, NJ
Rickson FR (1973) Review of glycogen plastid di€erentiation in
MuÈllerian body cells of Cecropia peltata. Ann N Y Acad Sci
210:104±114 Communicated by P. Pamilo