BlOTROPlCA 29(1): 84-92 1997

Influence of alfari on the Exploitation of Trees by a Leaf-Cutting

Heraldo L. Vasconcelos and Antonio B. Casimiro Coordenadoria de Pesquisas em Ecologia, lnstituto Nacional de Pesquisas da AmazBnia (INPA), Cx. Postal 478, 6901 1-970 Manaus, AM, Brazil

ABSTRACT The foraging activity of the leaf-cutting ant Ana laevigata in an early successional area near Manaus, Brazil, was monitored over a period of 18 mo. Four Cecropia were growing in that area and all were associated with the ant . Unoccupied trees (those in which there was no ant response to mechanical disturbance of the stem) were attacked by leaf-cutting ants more often than were trees occupied by A. alfari colonies. Cecropia ulei was the most frequently attacked species. However, leaves of C. ulei were the least preferred ones during assays in which detached leaves of the four Cecropia species were placed alongside the foraging trails of leaf- cutting ant colonies. C. ulei had the largest number of unoccupied trees compared to the other species. In addition, A. arfari colonies associated with C. ulei were smaller than those associated with C. distacbyu. These data suggest that A. alfari was influencing the selection of Cecropia species by A. laevigata, which harvested more from species that were less defended by the ants (with a higher proportion of unoccupied trees or trees hosting smaller ant colonies), although these were less preferred.

RESUMO A atividade forrageira da saliva Ana laevigata foi rnonitorada por 18 rneses em uma Area pr6xima a Manaus. Quatro esptcies de embaliba (Cecropia), associadas a formiga Aztecu ulfari, ocorriam nesta Area. Embalibas n5o ocupadas por A. alfari (aquelas em que nenhuma formiga emergiu do tronco da planta quando este foi sacudido) foram mais atacadas pelas salivas do que as embalibas ocupadas. Cecropia uki foi a esptcie mais atacada pelas salivas. Entretanto as folhas de C. uki foram menos preferidas pelas salivas em ensaios de campo nos quais folhas das quatro espkcies foram simultaneamente colocadas ao longo das trilhas das salivas. Urn nhmero maior de C. ulei nlo tinha colbnias de A. ufuri, comparativamente is outras espkcies. AItm disto, as colbnias de A. alfari em C. uki eram menores que aquelas de ao menos uma outra espkcie, C. distuchya. Estes dados sugerem que A. afuri afetou a escolha das espkcies de Cecropia por A. laevigatu, a qua1 atacou mais a esptcie que era menos defendida por A. uyuri (que tinha menos plantas colonizadas e que abrigava colBnias menores) memo que esta tinha folhas pouco palativeis.

Kty word: Amazonia; ant-plant interactions; Atta laevigata; Azteca alfari; Cecropia; herbivory.

TREESOF THE Cecropia (Cecropiaceae) are It has been suggested that Azteca also protect characteristic and important elements of many for- Cecropia against herbivory by leaf-cutting ants of est successional sites in the Neotropics. Most Cec- the genus Atta (Jolivet 1987, 1990), based on ob- ropia species are associated with ants, which live in servations that leaf-cutting ants rarely attack Cec- their hollow stems and feed on glycogen-rich, food ropia, in contrast to their attacks on surrounding bodies provided by the plant on a special organ at tree species, which are not associated with ants (Jo- the base of the petiole (Rickson 1971). Azteca spp. livet 1990). Several ant species are known, or are are the most common ant associates of Cecropia strongly suspected of protecting plants against leaf- (Wheeler 1942, Harada & Benson 1988, Longino cutting ants (Janzen 1967; Eberhard & ffiry 1991), particularly in more open habitats. In return 1974; Leston 1978; Jutsum et al. 1981; Cherrett for food and nesting space, some species of Azteca & Jutsum 1983; Dejean et al. 1992, 1995; Farji are known to protect their host-plants against en- Brener et al. 1992; Morawetz et al. 1992; Wetterer croaching vines (Janzen 1969, Schupp 1986, Da- 1994), and most of these species are similar to Rz- vidson et al. 1988) and Coleopteran herbivores teca of Cecropia in being arboreal and having ter- (Schupp 1986, Rocha & Bergallo 1992). ritorial habits. However, it is possible that the lower rate of leaf-cutting ant attack on Cecropia, as noted I Received 6 February 1995; revision accepted 11 Octo- by Jolivet (1990), is not associated with the pres- ber 1995. ence of defensive Azteca ants, but simply with the

84 Influence of Azteca on Leaf-cutting Ant Herbivory 85 lower palatability of Cecropia leaves compared to (height of the apical leaf attachment site) and re- other plants. This paper analyzes the effects Azteca ceived an uniquely numbered tag. Additional mea- have on herbivory by leaf-cutting ants on Cecropia. surements of height were conducted 8 and 16 mo later. Plant growth rates were calculated as: lnHl STUDY SITE AND SPECIES - InHO, where HO is tree height (m) at the begin- ning of the study and H1 is tree height 16 mo The study was conducted in a small (ca. 2 ha), later. abandoned farm, 7 km north of Manaus, Amazo- At intervals of approximately 14 days, over a to- ns, Brazil (3”05’S,60°00’W), where four species tal period of 18 mo, all trees were examined for of Cecropia were found growing sympatrically. The attack by leaf-cutting ants. At least three A. kzeui- site, on flat terrain at an elevation of 50 m, receives gata colonies were foraging in the study area, but an average of 2105 mm of rain annually (mean for we did not attempt to determine which trees were 1910 to 1979; Ribeiro & Adis 1984), which is attacked by which colony. A. kzeuigata builds un- seasonally distributed, with a distinct drier season derground foraging galleries radiating out from the between June and October. nest, and from the entrance to these, superficial This farm was used for several years to grow foraging trails extend to the plants being cut. This cassava and later as a plantation of Brazil nut trees. greatly increases the difficulty of determining the It was bulldozed in 1990, and has been abandoned foraging range and territories of individual colo- since then. In 1992, when this study was initiated, nies, as it is not always clear which entrance be- grasses and forbs had invaded and were dominating longs to which nest. the abandoned area. Trees, mostly Cecropia, oc- Trees attacked by A. kzeuigata were easy to rec- curred at relatively low densities. ognize as this species causes characteristic damage All four Cecropia species, Cecropia concolor Wild, by cutting through the petiole and removing whole Cecropia distachya Huber, Cecropia pulpurascens C. leaves, so that trees look as they have been pruned. C. Berg and Cecropia ulei Snethlage, were associ- In addition, scars are generally left on the basal part ated with a single ant species, Azteca alfari Emery of petioles, where the ants have attempted to cut (sensu stricto; Longino 1989). Previous studies in but have failed. A tree was recorded as being re-at- Manaus (Harada & Benson 1988) have also found tacked only after it had produced new leaves. A. alfari to be the most common associate of these four species, with only two of 61 plants being as- ASSAYS OF LEAF PREFERENCE.-kld assays were de- sociated with Azteca schimperi. The latter is easily signed to assess possible differences in preference to distinguished from A. a@i, both morphologically leaf-cutting ants among the leaves of the four Cec- (Longino 1991) and behaviourally (Benson 1988; ropia species. During the assays detached leaves A. Y. Harada, pers. comm.), as it builds external were placed alongside an active leaf-cutting ant for- carton nests. These were never found on plants of aging trail. Leaves of the four species (one per spe- our study site. Two leaf-cutting ant species, Atta cies) were presented simultaneously. The position higata Fr. Smith and Acromyrmex latireps nigro- of leaves from different species beside the trail was setosus Forel, were found in the study area, but A. chosen at random. Tracings were made of each leaf laticeps was never observed cutting Cecropia. before and 30 min after presentation. The total leaf area removed (cm2) was determined using a leaf METHODS area meter (Delta-T Devices, Burwell, Cambridge). Assays were performed with three A. kzeuigata col- LMF-CUTTINGANT FORAGING PATTEFWS.-AIIarea of onies and leaves from different trees were used for 0.39 ha, where evidence of leaf-cutting ant activity each assay. The leaves from the different species existed (as judged by the presence of damaged were collected from nearby plants of similar size. plants and foraging trails) was delimited. The area In an attempt to control for differences in leafage was divided in 10 x 10 m quadrats and all Cec- among leaves to be used in the assays, we always ropia growing within these quadrats were located collected the second or third leaf below the ter- and mapped. Only five new Cecropia emerged after minal meristem. This method of selection, how- this survey and these were not included in the anal- ever, presupposes that leaves from different Cecro- yses performed for this study. Each tree (including pia species develop at a similar rate, and this may 18 individuals that were protected from leaf-cut- not be the case. Nevertheless, all leaves collected ting ants within exclosure plots as part of another presented the coloration and the size of a fully ex- study; Vasconcelos 1994) was measured for height panded, mature leaf of its respective species. Data 86 Vasconcelos and Casimiro

(cm2 removed from each tree species) were trans- TABLE 1. Number of kaf-cutting ant attacks per indi- formed to In (x + 1) and were analyzed using a vidual tree, over a period of 18 mo, in jur randomized block analysis of variance, which con- Cecropia Ipecies and tbc percentage of tbc total sidered each replicated assay as a block. number of trccs attacked. Measurements of toughness and analyses of ni- Percent- trogen, fiber and water content were performed on Cecropia Attacks per tree age leaves of all four Cecropia species, leaf toughness species (mean 2 SE) attacked N

~~~ being determined with a penetrometer (Cherrett ulri 1.23 2 0.19 63.8 47 1968) which measured the force in grams required concolor 0.94 2 0.17 55.5 36 to push a standard needle point through the leaf. purpurascens 0.70 2 0.30 40.0 10 Nitrogen and crude fiber analyses followed the distacbya 0.51 2 0.14 27.3 55 A.O.A.C. methods (A.O.A.C. 1980).

OCCUPANCYBY AZTECA ANTS.-AU Cecropia were of 18 mo. C. ulei was the most frequently attacked examined for the presence of A. afari colonies, species, followed by C. concolor, C. purpurascem once in the middle of the study (in June 1993, or and C. distacbya (Table 1). The number of trees 8 mo afier it was initiated) and again 10 mo later attacked at least once differed significantly among when the study was completed. A tree was consid- Cecropia species and ranged from 63.8 percent in ered unoccupied if no ants emerged from its stem C. u&i to 27.3 percent in C. distacbya (x2 = 15.2, afier it had been shaken vigorously. Such mechan- df = 3, P = 0.002). The number of attacks per ical disturbance of trees occupied by A. arfari col- tree was also different among species (4144 = onies resulted in many workers emerging and 4.80, P = 0.003) and was significantly greater in swarming over the plant surfaces. This contrasts C. ulei than in C. distacbya (Tukey HSD multiple with the situation observed by Longino (1991) in comparisons test, P < 0.001). Costa Ria, who observed that when Cecropia trees inhabited by A. afari were shaken usually nothing ASSAYS OF LEAF PREFERENCE.-Although C. U&i WaS happened, suggesting that A. afari presents varia- the species most frequently attacked by leaf-cutting tion in behavior within different parts of its range. ants (Table 1) it was the least accepted one in the Alternatively or in addition, it is possible that the assays of leaf preference (Table 2). Leaves of C. ulei behavior of A. alfari changes with plant growth, as remained untouched in 7 of 10 assays, whereas in Longino worked with mature trees, whereas here the remaining assays only a few C. ulei leaf frag- we worked mostly with young trees and saplings. ments were cut. In contrast, leaves of both C. dis- To determine the size (total number of ant work- tacbya and C. purpurasrnuwere readily accepted by ers) of the A. afari colonies living in different Ccc- the ants in most assays. The mean area removed ropia species, 18 occupied trees (9 C. ulei and 9 C. from leaves of the latter species was almost 40 times distacbya) located near the main study area were greater than that removed from C. ulei (Table 2). collected. These trees were rapidly cut and placed Differences in leaf area removal by leaf-cutting in polythene bags containing ether to kill the ants. ants were not associated with any of the chemical The trees selected ranged in height from 0.9 to 2.7 and physical leaf characteristics measured (Table 2). m and none of them presented lateral branches. Leaves of C. dismbya and C. ulei had similar toughness despite large differences in the leaf area RESULTS removed by the ants, whereas those of C. distacbya and C. putpuracm, although equally acceptable to FORAGING rArrEms.-Cecropia trees attacked by A. the ants in the preference assays, were significantly hkatawere totally defoliated or left with only a different from each other in toughness, nitrogen few of their older leaves. In smaller trees (

TABLE 2. Area removed by kaf-rutting ants during assays of Ieafprefirence and characteristics of kaves ofjur Cecropia species.

Leaf characteristic Cecropia Area (cm2) Water Nitrogen Fiber species remove& contentb contenp contentd Toughnese uki 0.8 t 0.5 67.6 2 1.4 3.0 t 0.1 8.3 ? 0.9 16.7 ? 0.9 concolor 6.7 t 4.1 70.4 t 0.9 3.6 2 0.1 6.2 t 0.4 22.9 2 1.3 pulpurascens 30.3 t 9.1 68.7 ? 1.1 2.7 2 0.1 11.2 t 1.1 26.9 t 1.7 distachya 29.7 2 10.2 71.6 2 0.9 3.6 t 0.1 7.1 2 0.7 16.0 ? 1.0

a F3,27 = 4.42, P = 0.012. Percent fresh weight; F3,d3 = 2.73, P = 0.056. Percent dry weight; F3,16= 23.28, P < 0.001. Percent dry weight; F3,16 = 6.78, P = 0.004. Weight in grams to effect penetration of a needle point; F3,43= 16.10, P < 0.001.

a large one (3.4 m in height). All trees smaller than ropia species were highly significant (x2 = 20.27, 0.5 m had no ants. df = 3, P < 0.001). Trees of the first two species C. uki had the highest number of unoccupied tended to be smaller than those of the last two. trees (53.2%), followed by C. concolor (47.2%), C. Seventy-three percent of C. uki and 58.3 percent purpurascens (20%) and C. dirtachya (14.5%). Dif- of C concolor individuals were smaller than 1.5 m, ferences in frequency of ant occupancy among Cec- in contrast to 41.8 and 40 percent for C. d&tacLya

701 , , , 1 , , , , , I , C. ulei 1 80 50 50

40 40

30 30

20 20

10 10

n n 0.5 1.0 1.5 20 25 3.0 3.5 4.0 4.5 hO 0.5 1.0 1.5 20 25 3.0 3.5 4.0 4.5 5.0

60- C. purpurascens - 60- C. distachya -

20

10

~~ 0.5 LO 1.5 20 25 3.0 3.5 4.0 4.5 5.0 :I0.5 LO 1.5 20 25 3.0 3.5 40 45 5.0 TREE HEIGHT (m)

FIGURE 1. Size frequency distribution of trees of four Cecropia species with (slash filled bars) and without (black bars) colonies of Azteca alfari. 88 Vasconcelos and Casirniro

TABLE 3. Leafcutting ant attacks over a period of 18 mo be attacked more frequently than those of C. uki in four Cecropia species with and without col- and C. concolor. The opposite trend was found re- onies of Azteca alfari. garding occupied trees, those of C. distachya and C. purpurascens being attacked less frequently than Cecropia species With ants Without ants P-value those of the other two species (Table 3). Thirty-six percent of trees without A. alfari in Attacks per treea June 1993 were colonized within the next 10 mo, ulei 0.64 t 0.18 1.73 ? 0.27 0.002 whereas 9.4 percent of the trees with A. alfari lost concolor 0.68 t 0.21 1.24 ? 0.25 0.087 purpurascens 0.38 ? 0.26 2.00 t 0.00 0.025 their associated colony during the same period (Ta- distachya 0.21 t 0.08 2.25 ? 0.56 <0.001 ble 4). Colonization by A. alfari tended to be less Percentage of trees attackedb frequent in C. uki than in the remaining species ulei 45.4 (22) 80.8 (25) 0.031 and a greater number of occupied C. uki lost their concolor 42.1 (19) 70.6 (17) 0.086 colonies compared with other species (Table 4). purpurascens 25.0 (8) 100.0 (2) 0.133 The number of unoccupied trees that were colo- distach ya 14.9 (47) 100.0 (8) <0.001

~ ~~ nized after 10 mo was not significantly affected by a Mean 2 SE; P-values from one-way ANOVAs. whether the trees had been attacked or not by leaf- Sample sizes in parentheses; P-values from 2 X 2 Chi- cutting ants (Fisher exact probability test using data square, or Fisher exact probabiliry tesrs. from all species combined, P = 0.16). Similarly, there was no significant difference between at- and C. purpurascens, respectively (x2 = 11.05, df tacked and unattacked trees in the number of oc- = 3, P = 0.01 1). However, differences in tree size cupied trees that lost their associated colony (Fisher cannot fully explain the observed differences in ant exact probability test, P = 0.22). Of the 18 trees occupancy as these existed even for trees of similar that were protected from leaf-cutting ants, 42.9 size. While only 15 percent of C. distachya and 25 percent (N = 7) gained an A. alfari colony, whereas percent of C. purpurascens individuals between 0.5 9.1 percent (N = 11) lost their colonies. Analyses and 1.5 m in height had no ants, 60 percent of using combined data from unattacked and pro- those from C. uki and 68.8 percent of those from tected trees also revealed no significant differences C. concolor were without them. in colonization (x2 = 1.04, df = 1, P = 0.31) and Unoccupied trees were attacked by leaf-cutting survivorship (Fisher exact probability test, P = ants more often than were trees occupied by A. 0.22) of A. alfari colonies in these as compared to alfari colonies. Differences in the mean number of attacked trees. attacks between occupied and unoccupied trees No occupied tree in June 1993 (N = 96) was were significant for all species but C. concolor (Ta- dead 10 mo later whereas 25 percent (N = 52) of ble 3). A similar trend was observed regarding the the unoccupied trees had died (Fisher exact prob- proportion of attacked trees. From 71 to 100 per- ability test, P < 0.001). The mortality of unoc- cent of the unoccupied trees were attacked at least cupied Cecropia tended to be greater for those that once, whereas only 15 to 45 percent of the occu- were attacked by leaf-cutting ants but the differ- pied ones were attacked (Table 3). Unoccupied ence was not significant. Twenty-nine percent of trees of C. distadya and C. purpurascens tended to the attacked trees (N = 42) and 10 percent of the

~~~ TABLE 4. Changes in occupancy (percentage of unoccupied trees that were colonized and percentage of occupied trees that lost their associated colony) by A. alfari colonies in fiur Cecropia species over 10 mo.

Unoccupied trees that were colonized Occupied trees that lost their colony Cecropia species Totala Attacked Unattacked Totala Attacked Unattacked ufei 18.2 (22) 11.8 (17) 40.0 (5) 18.2 (22) 30.0 (10) 8.3 (12) concolor 45.4 (11) 28.6 (7) 75.0 (4) 5.3 (19) 0 (8) 9.1 (11) purpurascens 100.0 (2) 100.0 (2) ... b 12.5 (8) 0 (2) 16.7 (6) distachya 75.0 (4) 75.0 (4) ...b 6.4 (47) 14.3 (7) 5.0 (40) All species 35.9 (39) 30.0 (30) 55.6 (9) 9.4 (96) 14.8 (27) 7.2 (69) Number of observed trees in brackets. a Including trees attacked and trees not attacked by leaf-cutting ants. No data as all trees were attacked. Influence of Azteca on Leaf-cutting Ant Herbivory 89

3000 I I I I rn onies from C. uki were showing signs of senes- rn distachya 2500 cence, three of the nine examined colonies being 0 ulei queenless and some of the remaining ones having 2 5 2000 a relatively small brood. In contrast, all nine colo-

3i nies from C. distacbya had queens (one per colony) 1500 and plenty of brood. Six of the nine colonies from 3 each species had alates (males, females or both). 2 1000 z 500 GROWTHRATES OF CECROPU.-Different species of Cecropia presented significant differences in growth n 0.8 1.2 1.6 2.0 2.4 2.8 (Analysis of covariance, using number of leaf-cut- ting ant attacks as covariate: F3,149 = 7.10, < TREE HEIGHT (m) P 0.001). C. distachya tended to grow faster than the FIGURE 2. Relationship between tree height and the remaining species (Table 5), and this effect was in- size of the resident Azteca ayari colony in Cecropia dis- dependent of the number of times the tree was tachya and Cecropia uki. The estimated regression for C. attacked. Number of attacks had a marginally sig- distachya is: In CS = 5.75 + 0.79 TH, 9 = 0.80, P = 0.001 and for C. uki: In CS = 4.43 + 0.98 TH, 9 = nificant effect on tree growth (F1,,49 = 3.61, P = 0.60, P = 0.014, where CS is colony size as measured 0.059). Overall, trees attacked only once had a sim- by the total number of ant workers in the colony and ilar growth rate as those that were not attacked, TH is tree height (m). whereas trees attacked twice or more grew more slowly (Table 5). unattacked ones (N = 10) died (Fisher exact prob- DISCUSSION ability test, P = 0.21). PROTECTIONOF CECROPIABY AZTEa.-The results SIZEOF AZTECACOLONIES IN DIFFERENT CECROPU obtained here strongly suggest that A. alfari pro- srEcIEs.-Figure 2 shows the relationship between tects Cecropia against leaf-cutting ants. First, trees tree height and the size of the associated A. alfari without ants were attacked more frequently than colony for C. uki and C. distacbya. For both spe- those with ants. Second, the number of leaf-cutting cies, there was an increase in colony size (CS) with ant attacks on a given Cecropia species was not re- tree height (TH) (C. distacbya: In CS = 5.75 + lated to the palatability of its leaves, but to the 0.79 TH, 6 = 0.80, P = 0.001; C. uk: In CS = presence of A. alfari colonies. Most C. uki were 4.43 + 0.98 TH, 6 = 0.60, P = 0.014). The not occupied by A. a& which may explain why estimated regression lines did not differ in slope this species suffered a greater number of defolia- from each other (t-ratio = 0.22, P > 0.05; Zar tions, even though its leaves were less preferred by 1984), but they had significantly different inter- leaf-cutting ants than those of other Cecropia spe- cepts (t-ratio = 7.43, P < 0.001). A. a&% colonies cies. growing in C. distachya were larger than those Leaf-cutting ant attack on Cecropia was probably growing in C. uki, having from 2.1 to 3 times affected not only by the presence but also by the more workers, depending on tree height. Many col- size of the A. alfari colonies. C. distachya trees with

TABLE 5. Differential growth rates among Cecropia species and between trees subject to haf-cutting ant attack. Growth rate was cahkzted as: InHI-lnH0, where HO is tree height (m) at time zero and HI is tree height I6 mo hex

Cecropia species Not attackeda Attacked once Attacked twice or more uki 0.83 t 0.10 (21)b 0.95 2 0.20 (13) 0.52 2 0.12 (14) concolor 0.66 2 0.12 (20) 0.46 2 0.10 (7) 0.48 t 0.12 (8) puyurascenr 0.60 t 0.05 (6) 0.34 (1) 0.61 2 0.04 (3) distachya 1.04 2 0.06 (50) 1.01 2 0.06 (6) 0.84 2 0.14 (5) All species 0.89 2 0.05 (97) 0.81 2 0.11 (27) 0.58 2 0.07 (30) a Including individuals protected from leaf-cutting ants. Mean 2 SE. Sample sizes in parentheses. 90 Vasconcelos and Casirniro

A. alfari suffered a lower number of attacks than found no significant differences in colonization and those of C. ulei (Table 3). However, A. alfari col- survivorship of A. alfari colonies between attacked onies from C. distachya were larger than those from and unattacked trees (Table 4), we can not rule out C. ulei (Fig. 2), suggesting that larger A. alfari col- the possibility that leaf-cutting ant herbivory, by onies were more efficient in deterring leaf-cutting affecting Cecropia growth and performance, was ants from defoliating Cecropia. Similarly, studies the cause of the observed variations in ant occu- with (Rocha & Bergallo 1992) pancy. The analysis of the effects of leaf-cutting ant show that larger colonies are more efficient in find- herbivory on A. alfari colonization and survivor- ing and expelling herbivorous beetles from the ship was performed using data from all tree species leaves of its host-plant (Cecropia pacbystachya),and and tree sizes combined and therefore is a conser- that in consequence trees with larger colonies have vative analysis. However, if indeed leaf-cutting ants less herbivory than those with smaller colonies. In affected Azteca colonization one would expect to the ant-plant Maieta guianensis also, an inverse re- observe a lower rate of colonization on those un- lationship was found between the size of the resi- occupied trees that were attacked by leaf-cutting dent ant colony (as indicated by the number of ants more often. This was not the case as although workers patrolling the plant surface) and herbivore unoccupied C. distacbya and C. purpurascens suf- damage (Vasconcelos 1991). Casual observations of fered more attacks they were colonized by A. alfari diurnal attacks of A. laevigata on Cecropia showed more frequently than unoccupied C. ulei and C. that, when ants started to cut leaves from an oc- concolor. Also, similar variation in ant occupancy is cupied tree, patrolling A. alfari recruited their nest- reported by Harada (1982 and pers. comm.) who mates which scattered over the plant surface and worked in areas where the density of leaf-cutting attacked the leaf-cutters. The latter usually retreat- ant nests seems to be smaller than in the area de- ed and assembled at the base of the stem, starting scribed here, and therefore where the possible ef- to climb back into the tree crown as attacks by A. fects of leaf-cutting ants are expected to be smaller. alfari diminished. These observations suggest that Fifty percent of the C. ulei and 16 percent of the defoliation is probably avoided either when scouts C. concolor trees examined by Harada (1982) had of leaf-cutting ants are discovered and deterred be- workers but not queens, whereas none of the C. fore they can recruit; or when the number of A. distahya and C. purpuracens trees with A. alfari alfari workers largely exceeds the number of leaf- were queenless. This indicates that there is a greater cutters in the early stages of recruitment, as ants colony mortality in individuals of the former spe- from a large Atta nest would easily overwhelm cies, assuming that A. alfari colonies do not replace those from an A. alfari nest. In both cases the size their dead queens, which is probably the case as of the resident A. alfari colony would probably established colonies rarely have more than one matter. queen (Harada 1982, 1989). Studies in Costa Rica (Longino 1991) suggest Some species of Cecropia may be adapted to ac- that A. alfari is a poor defender of Cecropia com- quire their ants earlier than others or provide a pared with other Azteca species (some of the trees greater amount of resources, and this may explain with A. alfari being heavily infested with caterpil- the observed variations in occupancy by A. alfari. lars), whereas Andrade and Carauta (1982) work- In Cecropia the production of ant food bodies ing in southern Brazil question, on similar grounds, (Miillerian bodies and pearl bodies) declines with the existence of a between A. alfari and leaf age, and is greater in faster than in slower Cecropia. There is no information about the effec- growing trees (Davidson & Fisher 1991, Folgarait tiveness of other Azteca species against leaf-cutting & Davidson 1994). Coccids, which supply Azteca ants, but data presented here indicate that Cecropia with sugar-rich honeydew, are most often found in certainly benefits from its association with A alfari, the younger Cecropia internodes (Harada 1989) at least in sites like the one studied where other and may also be more abundant in a faster growing species do not occur. Repeated attacks by leaf-cut- tree. As A. alfari does not forage away from its ting ants reduced the growth of Cecropia trees (Ta- Cecropia, its is dependent on the food resources ble 5) and therefore reduced plant fitness. provided by its host-plant. A. alfari queens may therefore prefer to colonize faster growing trees, DIFFERENCESIN OCCUPANCY OF CECROPUBY ATE-whereas established colonies may perform better in i~.-TUhy the number of occupied trees and the these trees. The observations that C. distacbya, size of the resident A. alfari colonies differed be- which was the faster growing species, presented a tween Cecropia species is not clear. Although we greater number of occupied trees and that occupied Influence of Azteca on Leaf-cutting Ant Herbivory 91

trees were hosting bigger colonies are consistent ACKNOWLEDGMENTS with this view. In conclusion, and regardless of why A. a&ri We thank M. Cherrett for his support during the course established more readily in some Cecropia species, of this study and D. Davidson, A. Harada, R. Hutchings it is clear that A. a&ri plays an important role not and H. Fowler for reading and/or commenting on pre- vious versions of this manuscript. Manol and Maria Ines only in intra-/but also in inter-specific plant selec- Pereira Filho kindly performed the chemical analyses on tion by foraging leaf-cutting ants. Cecropia species Cecropia leaves and A. Harada identified the Azteca ants. of high palatability are likely to be poorly repre- Funds were provided by the Conselho Nacional de De- sented in the diet of leaf-cutting ant colonies if senvolvimento Cientifico e Tecnol6gico (CNPq grants these are well defended by ants, while species of 200785/90.7 and 521102/95.2) and by the Biological Dynamics of Forest Fragments Project (INPA/Smithson- lower palatability that are poorly defended may be ian Institution). This is publication number 145 of the very well represented. BDFFP Technical Series.

LITERATURE CITED

ANDRADE,J. C., AND J. I? l? CARAUTA.1982. The Cecropia-Azteca association: a case of mutualism? Biotropica 14: 15. A.O.A.C. 1980. Official methods of analysis. Association of Official Analytical Chemists, Washington, D.C. BENSON,W. W. 1988. Amazon ant-plants. In G. T. Prance and T. E. Lovejoy (Eds.). Amazonia, pp. 239-266. Pergamon Press, Oxford. CHEWTT, J. M. 1968. A simple penetrometer for measuring leaf toughness in feeding studies. J. Econ. Entomol. 61: 173617323. ,AND A. R. JUTSUM.1983. The effects of some ant species, especiallyAtta cephalotes L., Acromyrmex octospinosw (Reich) and Azteca sp. (: Formicidae) on citrus growing in Trinidad. In l? Jaisson (Ed.). Social in the tropics, pp. 155-163. University of Paris, Paris, France. DAVIDSON,D. W., AND B. L. FISHER.1991. Symbiosis of ants with Cecropia as a function of light regime. In C. Huxley and D. F. Culver (Eds.). Ant-plant interactions, pp. 289-309. Oxford University Press, Oxford, England. , J. J. LONGINO,AND R. R SNELLING.1988. Pruning of host plant neighbors by ants: an experimental approach. Ecology 69: 801-808. DEJEAN,A., I. OLMSTED,AND J. R. CAMAL.1992. Interaction between Atta cephalotes and arboreal ants in the Biosphere Reserve Sian Kaan (Quintana-Roo, Mexico): efficient protection of the trees. Sociobiol. 20: 57-76. --, AND R. R. SNELLING.1995. Tree-epiphyte-ant relationships in the low inundated forest of Sian Ka'an Biosphere Reserve, Quintana-Roo, Mexico. Biotropica 27: 57-70. EBERHARD,G., AND 0. KAFURY.1974. La ecologia de la hormiga Azteca trigona, una possible defensa contra las arrieras. Mem. 11, Congress0 de la Sociedade Colombiana de Entomologia, Cali, Colombia. FARJIBRENER, A. G., l? FOLGARAIT,AND J. PROTOMASTRO.1992. Associacion entre el arbusto Capparis retusa (Cap- paridaceae) y las hormigas Camponom slandw y Arromyrmex striam (Hymenoptera: Formicidae). Rev. Biol. Trop. 40: 341-344. FOLGAKAIT,l? J., AND D. W. DAVIDSON.1994. Antiherbivore defenses of myrmecophytic Cecropia under different light regimes. Oikos 71: 305-320. HARADA,A. Y. 1982. Contribuiggo ao conhecimento do ginero Azteca Forel, 1878 (Hymenoptera: Formicidae) e aspectos da interagzo com plantas do ginero Cecropia. M. Sc. Thesis, INPA e Universidade do Amazonas, Manaus, Brazil. . 1989. Estrutura de colBnias de &era ayari Emery (Hymenoptera, Formicidae, ) em plantas de Cecropia concolor (Moraceae). Revta. bras. Ent. 33: 169-182. , AND W. W. BENSON.1988. Espkies de Azteca alfari Emery (Hymenoptera, Formicidae) especializadas em Cecropia spp. (Moraceae): distribui6o geogrifica e considerag6es ecol6gicas. Revta. bras. Ent. 32: 423-435. JANZEN,D. H. 1967. Interaction of the bull's horn acacia (Acacia cornigera) with an ant inhabitant (Pseudomyrmex fmginea) in eastern Mexico. Kansas Univ. Sci. Bull. 47: 315-558. . 1969. Allelopathy by myrmecophytes: the ant Azreca as an allelopathic agent of Cecropia. Ecology 50: 147- 153. JOLIVET,l? 1987. Remarques sur la biocenose des Cecropia (Cecropiaceae). Biologie des Coelomera cheurolat avec les description d'une nouvelle espece du Bresil (Coleoptera, Chrysomelidae, Galerucinae). Bull. SOC.Linn. Lyon 56: 255-276. . 1990. Relative protection of Cecropia trees against leaf-cutting ants in tropical America. In R. K. Vander Meer, K. Jaffe and A. Cedeno (Eds.). Applied myrmecology: a world perspective, pp. 251-254. Westview Press, Boulder, Colorado. JUTSUM,A. R., J. M., CHEWIT AND M. FISHER.1981. Interactions between the fauna of citrus tree in Trinidad and the ants Atta cephalotes and Aweca sp. J. Appl. Ecol. 18: 187-195. LESTON,D. 1978. A Neotropical ant mosaic. Ann. Entomol. SOC.Am. 71: 649-653. 92 Vasconcelos and Casimiro

LONGINO,J. T. 1989. of the Cecropia-inhabiting ants in the Azteca arfari species group (Hymenoptera: Formicidae): evidence for two broadly sympatric species. Contributions in Science 1-16. . 1991. Azteca ants in Cecropia trees: taxonomy, colony structure, and behavior. In C. R. Huxley and D. F. Culver (Eds.). Ant-plant interactions, pp. 271-288. Oxford University Press, Oxford, England. MORAWETZ,W., M. HENZL,AND B. WALLNOFER.1992. Tree killing by herbicide producing ants for the establishment of pure Tococa occidentalis populations in the Peruvian Amazon. and Conservation 1: 19-33. RIBEIRO,M. N. G., AND J. his. 1984. Local rainfall variability: a potential bias for bioecological studies in central Amazon. Acta Arnazonica 14: 159-174. RICKSON,F. R. 1971. Glycogen plastids in Miillerian bodies of Cecropia peltata, a higher green plant. Science 173: 344-347. ROCHA,C. F. D., AND H. G. BERGALLO.1992. Bigger ant colonies reduce herbivory and herbivore residence time on leaves of an ant-plant: Azteca mueLleri vs. Coelomera ruficomis on Cecropia pacbystacLya. Oecologia 9 1 : 249- 252. SCHUPP,E. W. 1986. Azteca protection of Cecropia: ant occupation benefits juvenile trees. Oecologia 70 379-385. VASCONCELOS,H. L. 1991. Mutualism between Maieta guianenris Aubl., a myrmecophytic melastome, and one of its ant inhabitants: ant protection against insect herbivores. Oecologia 87: 295-298. . 1994. Interactions between leaf-cutting ants and forest regeneration in Amazonia. Ph.D. Thesis, University of Wales, Bangor, Wales. WETTERER,J. K. 1994. Attack by Paraponera clavata prevents herbivory by the leaf-cutting ant, Atta cqbaIotes. Biotropica 26: 462-465. WHEELER,W. M. 1942. Studies of Neotropical ant-plants and their ants. Bull. Mus. Comp. Zool., Harvard 90: 1- 262. ZAR,J. H. 1984. Biostatistical analysis. Prentice-Hall, Inc., Englewood Cliffs, New Jersey.