THE PERSISTENCE OF ANT-FOLLOWING BIRDS IN SMALL AMAZONIAN FOREST FRAGMENTS. Lee H. Harper (*) SUMMARY Uiòt-net captuneò <a£ a/uny ant-fallowing bi/idò weJie xf.onitoh.en. during xhe. isolation ofa centJial Amazonian faorest faagments ofa 1 ha [n = 5), 10 ha [n = 4) and 100 ka[n = 1). Post-isolation çaptu&eA ofa the, thn.ee obligate, ant-hollowing bind* VendAocinda meJvula, ViXhys albifywns, and Gymnopitkys Hufaigula decreased signifaicantty in all faragments. Post- isolation captures ofa faour facultative, ant-following spicieswere not significantly difafaft ent in faorest fragments o fa 1 ha and 10 ha, although two species decreased significantly in the. 100 ha fragment. ExpeJU.me.ntal introductions ofa obligatespecies into mall forest fragments tn the. absence and presence, o fa introduced Eexton bu/icheZli amy ant colonies resulted in significantly greater recaptures o fa introduced bi/j.ds when active amy ant colonies were present. Ofa the 105 bixdo introduced, 5S '55%) crossed 100-320 m od de dorested area and were recapture.d in continuous forest. INTRODUCTION The depauperate faunas of land bridge islands and continental forest isolates are well documented (Abbott, 1978; Diamond, 1975; Forman et al. , 1976; Gal 1 i et al . , 1977.. Leek, 1979; Terborgh, 197*»; Willis, 197^, 1979). The extinct ion of birds (Diamond, 1972; Terborgh, 197*0, mammals (Brown, 1971), and reptiles (Wilcox, 1978) have been related to the isolation area reduction associated with island formation. Some studies demonstrate that extinction probabilities may vary widely among species (Terborgh ε Winter, 1980; Willis, 1 972», 1979; Karr, 1982) and that faunas of smal ler i slands have higher extinction rates (Diamond, 1972, 1973; Terborgh, 197*»; Soule et al., 1979). However, stud ies of this type often are based on assumptions with regard to estimating the island's pre-isol ation fauna 1 community and the date of island formation. Faunal communities may vary over geo graphically short distances, and the potential for a sampl ing or founder effect in island systems is great. Extinctions can be cited erroneously for species that were not present (*) Biology Department,. St. Lawrence Univ., Canton, NY 13617· ACTA AMAZÔNICA, 19(Ün1co): 249-263. 1989. on the island at the time of isolation. These important aspects of island studies are often difficult to determine. Birds are vulnerable to many factors following isolation and area reduction \r) islands on forest fragments, including habitat alteration, increased abundance of preda^ tors, changes in resource levels, or increased abundance of competitors (Abbott, 1980)j The most well documented case of species extinctions on a land-bridge island is Barro Colorado Island in Panama. Both census data of the pre-isolat ion bi rd community and the date of island isolation are available. Since the istand was formed in 191*», the long term insularization effects described by Willis (197*») and Terborgh (197*») have resu 1 ted in the loss of at least *»5 of the original 208 bird species in about 50 years. Karr(1982) estimates that 50-60 forest bird species have been lost from Barro Colorado I sland. Both habitat succession and the increased abundance of certain mammalian predators may help explain many of the extinctions on Barro Colorado Island (Willis, 197*»', Terborgh & Winter, 1980). Although area is certainly an important factor which regulates species loss, both the size and quality of an island or forest fragment are critical ίη determining the rate and order of species disappearance (Diamond, 1978; Abbott, 1980). The relative quality of a forest fragment may be measured by the quantification of some critical resource for the organism under study. The army ant-following guild of birds is a system for which available resource levels may be estimated by monitoring the number of colonies of the army ant Eciton burchelli in the forest fragment. The regular swarming cycle of these army ants provides a dependable source of flushed insect prey for several- species of antbirds (Formi car i idae), woodcreepers (Dendrocol apt idae) and other birds (Willis & On_i_ k i, I978). Although most families of birds show mixed susceptibi1ity to extinction (Te£ borgh & Winter, 1980), antbirds have been cited as extinction prone where the area of continuous forest has been reduced, such as at Barro Colorado Island (Terborgh, 197*1; Wil lis, 197*») and in tropical forest isolates (Leek, 1979; WM lis, ,1979). I selected four species of antbirds and six species of woodcreepers (Table 1) to examine the persistence of individuals in forest fragments of varying area iη conjunct ion with the INPA-WWF Biological Dynamics of Forest Fragments project in Manaus, Amazonas, Brazil. ΑΠ of these species follow army ants at least facultatively, and three can be considered obligate and followers (Oniki & Willis, 1972; Willis & Oniki, 1978; Willis, 1977; Willis, 1983; Harper, unpubl. data). Often thought to be sedentary and reluctant to cross open areas, antbirds and other ant-following birds are ideal for studying the process of species disappearance in remnant forest fragments. They are common in the forest understory, easily captured and color-banded, and can be approached closely for observation. The birds attend ant swarms for several hours each day and obi igate species gain a significant portion of their diet by foraging in this manner. Foraging behavior within the guild is restricted by a weight- and s i ze-structured dorr: nance hierarchy wh i ch may exclude smaller species from the most productive part of an ant raid (Wi11 is & Oniki, J 978). There- were four objectives to- roy study-; 1) te -deter-mine the effects of area and 25O Harper resource reduction on the persistence of these ten Bird species in isolated forest fragments, 2) to perform experimental introductions of the species into an isolated 1 ha and an isolated 10 ha fragment to rule out the possibility of a sample effect bias in fragment formation, 3) to determine if post-isolation changes occur in the availability of insect prey in continuous forest and forest fragments in the presence or absence of army ants, and 4) to experimentally and indirectly increase resource availability in a 10 ha forest fragment by introducing colonies of the army and E. burchelli to determine the effect of habitat qual i ty on the pers i stence of ant-foUowi ng bi rds i η forest fragments . I predicted that the mist-net captures of obligate ant-following species would decrease significantly in forest fragments of 1 ha, 1C ha and 100 ha, that experimental introductions would demonstrate this to be real and not a sample effect, and leaf litter insect resources would not significantly decline within the forest fragment. I further predicted that introduced obligate ant-following birds would be recaptured iη the fragment as long as it contained active colonies of the army ant E. burchelli. If individuals of a species disappeared when army ants were present, the greatest dec 1ine in numbers would be inversely correlated with the position of that species in the dominance hierarchy. However, for facultative ant-following species, I predicted no differences would be observed between pre- and post-isolation capture rates in the 1 ha, 10 ha 100 ha forest fragments. METHODS The presence or absence of the four antbird and six woodcreeper species was exami ned in five forest fragments of I ha, four forest fragments of 10 ha, one partially iso_ lated 100 ha fragment, and in continuous forest (Table 2). These data were obtained through mist-netting, and visual and acoustical censuses. All fragments were mist-netted once every three to four weeks. Approximately 130 mist-net hours (nhrs) were logged in a single banding day for each net line by opening 16 mist-nets at dawn and starting to close these nets at )k00 hrs. This sampling program is described Tn detail Bierregaard et al. (this volume). In addition, the 10 ha fragment used for the experimental intro­ ductions was saturated with 52 mist-nets to capture any introduced birds or colonizing individuals which remained in the forest fragment. These data were analyzed and tested with Friedman two way ANOVAs to compare pre- and post-isolat ion capture rates for 1 ha (N = 5, Table 3) and 10 ha (N = k, Table k) fragments. Because the sample size was only one for the 100 ha fragment, captures per 100 nhrs for pre-isolation banding days were compared to post-isolation captures with Mann-Whitney U tests (Table 5). The predicted decrease in capture rates for obi igate ant followers required a one-tailed test, while the lack of directional a priori^ predictions for facultative ant followers required a two-tailed test. A total of 6| Pithys albifrons, 31 Gymnopithys rufigula and 12 Dendrocincla merula were experimentally introduced into the isolated 1 ha and 10 ha forest fragments (Table 6). These birds were captured with four to eight mist-nets strung along narrow trails in front of raiding army ant swarms, within 1.5 km of the release site. The nets were left open for approximately 30 minutes with care taken to prevent birds being killed by the ants. Birds were held temporarily in cloth bags for transport and release in the isolated fragments. To determine if the introduced birds remained in the isolated fragments, daily census observations and mist-netting every two or three days were per­ formed until the army ant colony abandoned the fragment. Insects were sampled in the presence and absence of swarming army ants with nine 50 χ 50 cm clear plexiglas plates and eighteen 15 cm diameter petri dishes covered with Tanglefoot, and equally spaced along a 9 m transect. Traps were left open for twenty minutes before the insects were removed. The insects were stored in 70¾ alcohol and later dried in an oven for **8 hrs and weighed on a precision balance. Subsamples of these insects were identified at least to family. A minimum of five trap transects were collected in each of three isolated 1 ha fragments, three isolated 10 ha fragments, and the continuous forest. Five transects were collected in front of army ant raids in continuous forest and another five in isolated 10 ha fragments containing army ants, including the exper_i_ mentally introduced army ant colonies. AH samples were collected between 0900 and 1300 hrs on days without rain. Samples were statistically compared with two-tailed Mann- Whi tney U tests.. Four colonies of the army ant E. burchelli were captured and individually released in the center of the experimental 10 ha fragment (Table 7)· Army ant colony introductions involved the capture of the queen during nocturnal nomadic migrations and isolating her in 5 x 3 x 3 cm wire holding cage with 2 mm mesh. This cage was suspended on a structure of palm fronds and sticks in a plastic bucket of approximately 75 liters volume and placed near the new bivouac site. If the ant colony formed a bivouac cluster around the queen during the night, the bucket was sealed before dawn, weighed, moved to the 10 ha fragment and released. The bucket was later weighed again to provide the live weight of the ant colony. Subsamples of live ants were collected, weighed and counted to calcu late colony size. Following the release of each of the three largest E. burchelli col£ nies, individuals of the three obligate ant-following bird species also were released into the isolated fragment. None of these individuals had been used in earlier experi ments . All data analysis and statistical tests were performed on a Compaq Plus computer using the statistical software package "Statpak" from Northwest Analytical. RESULTS The responses of ant-following birds to forest fragmentation differed markedly (Figure 1, Tables 3, and 5). Figure 1 plots mean post-isolation captures per 100 nhrs against mean pre-isolation captures per 100 nhrs for seven species of ant-fol lowing birds. The diagonal line represents equal pre- and post-isolation capture rates.- Symbols ρ lot led abe/e the line indicate higher capture rates i η forest fragments than in cont inuous forest; symbols below the line indicate lower capture rates in forest fragments. Standard errors have not been included in Figure 1 but are given in the tables. Lines have been drawn between the symbols to facilitate species identification. Capture rates for the facultative ant-followinq species did not siginicantly change in the 1 ha (n = 5, Table 3) and 1ΰ ha (n = k, Table k} fragments following forest fragmentation (Friedman two way ANOVA, NS). The three 1argest ant-following woodcreepers Oendrocolaptes picumnus, Dendrocolaptes certhia, and Hylexetastes perrotii were rarely captured in continuous forest, al though thi s pattern changed somewhat i η fragmented forest. Captures of H. perrotii decreased in the 1 ha fragments, increased in the 1 Oha fragments, and remained unchanged in the 100 ha fragment. The woodcreeper D. certhia was captured more frequently in the 1 ha fragments after isolation. These three woodcreepers, as well as D. merula were observed or captured after having temporarily colonized or invaded isolated 1 ha or 10 ha forest fragments 75-150 m from the adjacent continuous forest. Individuals of D. certhia were present in k of 5 (80¾) 1 ha fragments after a year of isolation, and continued to be captured occasionally in one 1 ha fragment years after isolation. Pre- and post-isol at ion capture rates for these species were low (Tables 3» A and 5), and were therefore not included in Figure 1. Capture rates of Dendrocincla fuliginosa decreased in the 1 ha fragments but re mained stable in the larger fragments, and this species continued to persist in forest fragments of all sizes. Xiphorhynchus pardalotus was captured cons iderably more of ten in the I ha and 10 ha fragments following isolation, but -not to a significant degree. Captures of the two antbirds, Hylophylax poecilinota and Percnostola rufifrons, varied between the two sizes of small fragments. While captures of H. poeci1inota decreased in the 1 ha fragments, th i s species increased in the post-i sol at ion data of the IOha fragments. P. rufifrons showed the opposite pattern, and increased in the 1 ha fragments while de creasing slightly in the 10 ha fragments. In the 100 ha fragments (Table 5), captures of D. fuliginosa and H. perrotii re mained stable, while the remaining four facultative ant-following species were captured less frequently after complete isolation, P. rufifrons and H. poecilinota to a signifi cant degree (two tailed Mann-Whitney U test, Ρ <.025). Capture rates of all three obligate ant-following species,D.merula , P. albifrons, and G. rufigula decreased significantly in fragments of al1 sizes. Although these species were common in continuous forest, their capture rates approached or reached zero after fragmentation. Actual post-isolation capture rates were higher in the 100 ha fragment than in the fragments of 1 ha or 10 ha. None of the 1 ha nor the 10 ha fragments con­ tinuously supported active army ant colonies following isolation. The 10-0 ha fragment was observed to support as many as three army ant colonies after fragmentation, al though there were periods when none of these colonies swarmed on a given day. The dry weight of insect samples collected in the presence of swarming army ants was significantly higher than those samples collected in the absence of ants of ants con tinuou* forest and forest fragments (Figure 2, two—tailed Mann-Whitney U Test, P<.61). Insect sample weights were not different between the two habitat types in the presence or absence or army ants. In the absence of raiding army ants, the level of availabl$ insects was 6-7 times lower than in the presence of an army ant swarm raid. Army ant colonies raiding in isolated forest fragments flushed slightly lower levels of insect prey compared to continuous forest, but not to a significant degree. A subsample of th* insects collected in sticky traps revealed 60 fami 1 ies in l^orders. The ma jor i ty of thes$ insects were orthoptera (crickets, grasshoppers) and dicoptera (cockroaches), which are prime prey items bf ant-following birds. The experimental i nt roduct ions of obi igate ant-fo I lowi ng b i rds i nto forest fragments demonstrated that birds quickly disappeared from fragments not conta ining army ants. Only one individual of G. rufigula remained in the isolated 10 ha fragment 48 hrs after the birds were released (Table 6). However, when army ant colonies were captured and indi vidually introduced into this same 10 ha fragment (Table 7), six introduced G. rufigula and one P. albifrons were captured foraging at the army ant swarms one week after rel ease^ The total number of within fragment captures summed over all three species was signifi cantly higher in the presence of the introduced army ants £han in their absence (Table 6, G-test, P<.025). This was due to the higher number of captures of G.rufigula, which in itself was a significant increase (Table 6, G-test, P<.025). No introduced individuals of D. merula were observed or captured in the mist-nets. The null hypothesis that capture rates for each species would be equal in the presence and absence of army ants is re jected for G. rufigula. Although the disappearance of the subordinate P. albifrons was predicted, the disappearance of D. merula was not. Although no specific effort was made to search the continuous forest for the birds which disappeared from the forest fragments, 55¾ of the total (58 of 105) birds intro-< duced into the forest fragments were later recaptured in the continuous forest (Table 8). The lowest number of individuals recaptured was for G. rufigula at 39¾ (12 of 32), whereas 62¾ of introduced P. albifrons (38 of 61) and 62¾ of the introduced D. merula (8 of 13) were later recaptured in the continuous forest. DISCUSSION Many of the woodcreeper species did not remain restricted to these small isolated; forest fragments, but rather moved in and out across the deforested areas. Someof these1 woodcreepers, notably H. perrotii, D. picumnus, D. certhia, and D. merula were tempo rarily colonized or invaded isolated 1 ha or 10 ha forest fragments. These 1 ha and lOha fragments were 75-150 m from the adjacent continuous forest, and these colonizat ions mairi tained low capture rates of these species in small forest fragments. In this manner, D. certhia may be able to persist in small forest fragments close to a confinuous forest source area, although it would disappear from more remote fragments. This species eventu ally disappeared from the much larger and more isolated Barro Colarado Island after 50 ^ears (Willis* 197¾)- The Large woodcreepers often fly ax heights above, few© meter* and their actual abundance as measured with mist-net data may have been underestimated. Other species, such as X. pardalotus, are facultative ant followers that forage more frequently with mixed species flocks than ant-fol1owings flocks, and were frequently captured in mist-nets. The increase in capture rates observed for X. pardalotus were most likely due to the influx or crowding of individuals into the forest fragment from the surrounding areas of deforestation. This species was the only one to increase sub­ stantially in both the 1 ha and 10 ha forest fragments, where it appeared to persist quite wel1. P. rufifrons is frequently encountered along forest edges and in second growth, possibly explaining the observed capture rate increase in the 1 ha fragments and itssi_g_ nificant decrease in the interior of the 100 ha fragment. This was probably true of H. poecilinota also, which increased in the 10 ha fragment, an area siightly larger i ts home (Willis, 1982). Banding recoveries indicate that indiνiduals of H.poeci1inota move between isolated fragments, at times traveling in mated pairs. I η th i s manner, captures remained high. A similar explanation may be involved for X. pardalotus, although recovery data was insufficient. As predicted, the three obligate ant-following species, D. merula, P. albifrons, and G. rufigula, virtually disappeared from all the forest fragments regardless of size. Although army ant colonies were observed in the 100 ha fragment, the irregular swarming cycle of these colonies resulted in periods with no active ant swarm available for the birds. Interestingly, the capture rates of these obligate ant-following species re­ mained very slightly above zero, indicating rare colonization of the the forest fragment by birds in search of ants. The recovery data from the experimental introduction of birds into the isolated 1 ha and 10 ha fragments (Table 6) revealed only one G. rufigula recaptured in the fragment in the absence of ants. This single recapture occurred less than 48 hrs after the bird was released. No individuals of P. albifrons were captured in either the 1 ha or the 10 ha fragment. It was not surprising that ant-following birds did not remain in areas as small as 1 ha or 10 ha, as this size was considerably below the area of their home range. The interesting result was the tendency and speed wi th which the birds left the fragments. In one case, an individuals P. albifrons was recaptured at theoriginal continuous forest capture site only 1.75 hrs after it was released in a a 10 ha fragmentwi thout arm/ants. All but one of the introduced individuals disappeared from the fragment in less than 48 hrs in the absence of ants. These birds left the isolated forest fragments and crossed the deforested area into the adjacent forest. It was unlikely that any of the introduced birds perished in the isolated fragments, and 55¾ were recaptured later iη the continuous forest. In addition to the recaptures of introduced birds in the presence of introduced army ants, two G. rufigula and one D. merula that entered the fragment fr©m the sur­ rounding continuous forest were captured foraging at the ant swarms. The number of colo nizing individuals undoubtedly was higher; several other individuals we re obse rved f o rag_ ing with the ants, but were not captured nor identified by *4>e observation of color bands. Efforts to stabilize the ant colony within the 10 ha forest fragment were unsuc cessful, and the ants wandered out of the fragment eight days after release. Al though the ant-fi1 lowing birds appeared to leave the fragment following the departure of the ants, an examination of banding recoveries indicated that individuals of G. rufigula tempora­ rily colonized this 10 ha fragment at least seven times, often in the absence of army ants. These observations suggest that ant-following birds enter forest fragments to search for active army ant colonies, and do not simply follow raiding columns across de_ forested areas. Unfortunately, the ephemera! nature of these colonization events makes an estimate of their frequency difficult. All of the birds which entered this 10 ha forest fragment, as well as those which were introduced and later recaptured outside, crissed the deforested area which sepa­ rates the fragment from the continuous forest. This deforested area was between 100 m and 150 m in width. Those individuals which left the isolated 1 ha fragment crossed a distance of at least 320 m of cleared area to reach the continuous forest (Table 8). It is interesting to note that "at its nearest point, BCI [Barro Colorado Island] is [ap­ proximately] 200 meters from a peninsular extension" of the ma inland forest (Karr,1982). Although areas of deforestation represent a less severe dispersal barrier than open water, understory ant-following birds can cross significant open areas between forest fragments and continuous forest. Although actual levels of insects in the forest fragments were not different from the continuous forest (Figure 2), the lack of an active E. burchel1i colony and the ac­ companying decrease in food avai Iabi I i ty resul ted i η the d i sappearance of the ant-fol lowing birds. The immediate major effect of isolation was the sample effect of whether army ants were present, and not any changes in microclimate that may have affected the abso­ lute level of leaf litter insect abundance and the survival potential of ant-following birds. The capture and release of colonies of the army and E. burchelli into forest fragments raised the relative quality of the fragment to a sufficient level to at least temporarily support the ant-following bird community. Although the absolute levels of insect resources were not altered, the introduced army ants and their foraging raids made these resources available to the ant-following birds. The enrichment of a forest fragment through the introduction of army ant colonies mat help maintain populations of ant-following birds as long as army colonies in the fragment. Large areas and several army ant colonies would be necessary to preserve the ant-following bird community on a long term basis, and habitat suitability, resource availability and predation risk should be assessed carefully to avoid the failureof the introductions (Morton, 1978). Small forest fragments, 1 ike those examined here,hold 1 ittle conservation value; the nomadic behavior of army ants make it likely that a colony will wander out of the reserve boundaries, and the army ants or ant-following birds might .re_ duce or eliminate populations of other organisms before leaving or dyingout themselves. Large forest fragments, and the maintenance of forested corridors and short separation distances among them, would reduce the risk of species disappearance and community decay. However, the ability to capture and release intact army ant colonies and ant-following bands. Efforts to stabilize the ant colony within the 10 ha forest fragment were unsuç cessful, and the ants wandered out of the fragment eight days after release. Al though the ant-fi1 lowing birds appeared to leave the fragment following the departure of the ants, an examination of banding recoveries indicated that individuals of G. rufigula tempora­ rily colonized this 10 ha fragment at least seven times, often in the absence of army ants. These observations suggest that ant-following birds enter forest fragments to search for active army ant colonies, and do not simply follow raiding columns across de forested areas. Unfortunately, the ephemeral nature of these colonization events makes an estimate of their frequency difficult. All of the birds which entered this 10 ha forest fragment, as well as those which were introduced and later recaptured outside, crissed the deforested area which sepa­ rates the fragment from the continuous forest. This deforested area was between 100 m and 150 m in width. Those individuals which left the isolated 1 ha fragment crossed a distance of at least 320 m of cleared area to reach the continuous forest (Table 8). It is interesting to note that "at its nearest point, BCI [Barro Colorado Island] is [ap­ proximately] 200 meters from a peninsular extension" of the mainland forest (Karr,1982). Although areas of deforestation represent a less severe dispersal barrier than open water, understory ant-following birds can cross significant open areas between forest fragments and continuous forest. Although actual levels of insects in the forest fragments were not different from the continuous forest (Figure 2), the lack of an active E. burchelli colony and the ac­ companying decrease in food ava i 1 abi I i ty resul ted in the d i sappearance of'the ant-foll owing birds. The immediate major effect of isolation was the sample effect of whether army ants were present, and not any changes in microclimate that may have affected the abso­ lute level of leaf litter insect abundance and the survival potential of ant-following birds. The capture and release of colonies of the army and E. burchelli into forest fragments raised the relative quality of the fragment to a sufficient level to at least temporarily support the ant-following bird community. Although the absolute levels of insect resources were not altered, the introduced army ants and their foraging raids made these resources available to the ant-following birds. The enrichment of a forest fragment through the introduction of army ant colonies mat help maintain populations of ant-following birds as long as army colonies in the fragment. Large areas and several army ant colonies would be necessary to preserve the ant-following bird community on a long term basis, and habitat suitability, resource availability and predation risk should be assessed carefully to avoid the failureof the introductions (Morton, 1978). Small forest fragments, I ike those examined here ,hol d 1 i tt 1 e conservation value; the nomadic behavior of army ants make it likely that a colony will wander out of the reserve boundaries, and the army ants or ant-following birds might ,re_ duce or eliminate populations of other organisms before leaving or dying out themselves. Large forest fragments, and the maintenance of forested corridors and short separation distances among them, would reduce the risk of species disappearance and community decay. However, the ability to capture and release intact army ant colonies and ant-following birds holds promine for, the management of these communities in large forest fragments ι*ό biological reserves, and illustrates the potential for re-establishing the ant-following bird community in areas where it has been extirpated. ACKNOWLEDGMENTS This research was supported by the binational INPA/WWF-US Biological Dynamics of Forest Fragments project. Additional funds were provided by the State University of New York at Albany and Kenneth P. Able. I would especially like to thank Thomas E. Lovejoy and Richard 0. Bierregaard for their support. Thanks also to Norman Penny and Francisco Peralta who helped with the insect analyses and i dent i f ication ; George Powel 1 , Bert Klein, Carlos Quintela and Antonio Cabral who assisted with the mist-netting and provided hours of helpful discussion; and Luiz Raimundo Queiroz who assisted with the capture of the army ant colonies. This is publication number 39 in the Minimum Critical Size of Ecosystems Techni­ cal Series. RESUMO Capturas com sie.de, de pássaros seguidores de formigas-de-correição foram monitora das durante o isolamento de fragmentos de mata da Amazônia Central de lha {n = 5), 10 ha [η ~ 4) e 100 ha [n = 1). Capturas pós-isolamento de. três espécies de pássaros seguido^ fies de. formiga Vendrocincla merula, Pithys albifrons, e Gymnopithys rufigula diminuiram significativamente em todos os fragmentos. Capturas pos-isolamento de. quatro espécies facultativas seguijdoras de formigas não foram significativamente diferentes em fragmen tos de mata de 1 ha e 10 ha, embora duas espécies diminuíssem consideravelmente no farag_ \nento de 100 hs. Introduções experimentais de espécies obrigatórias em pequenos frao_ Centos na ausência e presença de colônias introduzidas de formiga-de-correição Eciton burchelli resultaram em significativamente maiores taxas de recapturas de pássaros in troduzidos quando as colônias ativas de formigas estavam presentes. Vos 105 pássaros introduzidos, 58 '55% atravessaram cerca de 100-320 m de mata devastada e foram recaptu rados na mata contínua.