Life above ground: ecology of Anolis fuscoauratus in the Amazon rain forest, and comparisons with its nearest relatives

142 Life above ground: ecology of Anolis fuscoauratus in the Amazon rain forest, and comparisons with its nearest relatives Laurie J. Vitt, Teresa Cristina S. Avila-Pires, Peter A. Zani, Shawn S. Sartorius, and Maria Cristina Espósito Abstract: The polychrotid lizard Anolis fuscoauratus was studied at six localities in the Ecuadorian and Brazilian Amazon from 1994 to 1999. Throughout the Amazon, A. fuscoauratus occurs in forested habitats, is arboreal on tree trunks, limbs, and branches as well as vines, has a body temperature (T b ) of 28.7 ± 0.2 C (mean ± SE) while active, maintains T b slightly above ambient temperatures, avoids direct sunlight during most of the day, and feeds primarily on a combination of orthopterans (20.62% by volume), spiders (16.7%), homopterans (10.62%), and insect larvae (10.35%). Despite detectable geographic variation in adult body size and diets, general ecological attributes are similar among populations across the Amazon region even though the number of sympatric Anolis species as well as the total number of lizard species vary among sites. Overall ecological similarity likely reflects the fact that there is little evolutionary divergence among populations. Comparisons between A. fuscoauratus and its three closest relatives, A. humilis and A. limifrons of Central America and A. trachyderma of the Amazon, reveal some similarities. All four species maintain relatively low T b while active. Anolis fuscoauratus and A. limifrons are ecologically and morphologically similar but A. fuscoauratus is larger. Anolis humilis and A. trachyderma are more similar to each other ecologically than they are to their respective sympatric congeners. Anolis humilis is smaller than and morphologically dissimilar to A. trachyderma. The Amazonian and Central American species pairs do not comprise each other s closest relatives, indicating that similar ecomorphs have evolved independently in the Amazonian and Central American rain forests. Résumé : Nous avons étudié la biologie d un lézard polychrotidé, Anolis fuscoauratus, à six localités de l Amazonie équatorienne et brésilienne, de 1994 à 1999. Dans toute l Amazonie, ce lézard vit dans les zones forestières et est arboricole sur les troncs, les branches et les rameaux, ainsi que sur les lianes; sa température corporelle (T b ) est de 28,7 ± 0,2 C (moyenne ± erreur type) lorsqu il est actif, toujours légèrement au-dessus de la température ambiante; il évite la lumière directe du soleil pendant presque toute la journée et il se nourrit principalement d un mélange d orthoptères (20,62 % en volume), d araignées (16,7 %), d homoptères (10,62 %) et de larves d insectes (10,35 %). Bien qu il y ait une variation géographique détectable de la taille des adultes et du régime alimentaire, les caractéristiques écologiques générales sont semblables chez toutes les populations d Amazonie, en dépit du fait que le nombre d espèces sympatriques d Anolis et le nombre total d espèces de lézards varient d un site à l autre. La similarité écologique générale reflète probablement le faible degré de divergence évolutive entre les populations. La comparaison d A. fuscoauratus aux trois espèces qui lui sont le plus apparentées, A. humilis et A. limifrons de l Amérique centrale et A. trachyderma de l Amazonie, a mis en lumière des similarités entre ces espèces. Les quatre espèces maintiennent assez basse leur température T b lorsqu elles sont actives. Anolis fuscoauratus et A. limifrons sont semblables écologiquement et morphologiquement, mais A. fuscoauratus est de taille plus grande. Anolis humilis et A. trachyderma sont plus semblables l un à l autre écologiquement qu à toute autre espèce sympatrique d Anolis. Anolis humilis est plus petit qu A. trachyderma et la morphologie des deux espèces est différente. Les paires d espèces de l Amazonie et de l Amérique centrale n incluent pas leurs espèces respectives les plus apparentées, ce qui indique que des écomorphes semblables ont évolué indépendamment en Amazonie et en Amérique centrale. [Traduit par la Rédaction] Vitt et al. 156 Received 20 June 2002. Accepted 29 November 2002. Published on the NRC Research Press Web site at http://cjz.nrc.ca on 21 February 2003. L.J. Vitt. 1 Sam Noble Oklahoma Museum of Natural History and Department of Zoology, University of Oklahoma, 2401 Chautauqua Avenue, Norman, OK 73072, U.S.A. T.C.S. Avila-Pires and M.C. Espósito. Departamento de Zoologia, Museu Paraense Emílio Goeldi, Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) Ministério da Ci ncia e Tecnologia, Caixa Postal 399, 66017-970 Belém, Pará, Brazil. P.A. Zani. Department of Biology, University of Oregon, Eugene, OR 97403, U.S.A. S.S. Sartorius. Department of Zoology and Oklahoma Museum of Natural History, University of Oklahoma, Norman, OK 73019, U.S.A. 1 Corresponding author (e-mail: vitt@ou.edu). Can. J. Zool. 81: 142 156 (2003) doi: 10.1139/Z02-230

Vitt et al. 143 Introduction Among the most common comments at annual meetings of professional scientific societies (e.g., American Society of Ichthyologists and Herpetologists, Herpetologists League, Society for the Study of Amphibians and Reptiles) and conservation-oriented fact-finding meetings (e.g., National Research Council s Irvine discussion of disappearing amphibians and the Atlanta meeting of the Partners in Amphibian and Reptile Conservation) is that basic natural-history and ecological data are lacking for most species of reptiles and amphibians (see also Greene 1986, 1994); yet these data are essential as a backdrop against which to measure the effects of environmental change on species. Natural-history data are also critical for development of realistic speciesand habitat-management strategies. Considering the remarkable insights that phylogenetic analyses of behavioral, physiological, and ecological data have produced (e.g., Garland and Carter 1994; Martins 1994; Losos 1995), species-based natural-history and ecology studies are increasingly important because they provide data necessary to track the evolution of biologically important phenomena. We consider such studies not only justified but absolutely necessary and timely, given the current rate of habitat and species loss throughout the world. For some of the potentially most interesting habitats, we will never even have accurate species lists or detailed natural-history data on individual species. The Amazon rain forest is a perfect example of a habitat that is rapidly becoming fragmented as humans convert forest to farm and pastureland (e.g., Fearnside and Ferraz 1995). With this in mind, we describe the ecology of the most common Amazonian anole, Anolis fuscoauratus. Only4of10Anolis species occurring in Amazonia have distributions that cover or exceed the entire basin: A. fuscoauratus, A. nitens (with five described subspecies; but see Glor et al. 2001), A. ortonii, and A. punctatus (Avila-Pires 1995). Of these, A. fuscoauratus is the smallest in body size, by far the most abundant, and the only one that appears to occur virtually everywhere as long as forest is intact. Anolis fuscoauratus is often found on low vegetation and vines but also uses tree trunks and climbs well into the tree canopy (Beebe 1944; Dixon and Soini 1975; Duellman 1978, 1987; Avila-Pires 1995; Vitt and Zani 1996c). It is a member of the clade containing the Central American anoles A. humilis, A. limifrons, and the Amazonian anole A. trachyderma (Glor et al. 2001). Both A. humilis and A. limifrons have been studied extensively (e.g., Sexton et al. 1963, 1972; Talbot 1979; Andrews and Rand 1982; Andrews et al. 1983; Guyer 1988a, 1988b, 1994; Andrews and Wright 1994; Parmalee and Guyer 1995) and a recent study describes the ecology of A. trachyderma (Vitt et al. 2002). Superficially, A. fuscoauratus appears similar in morphology to A. limifrons of Central America and to use many of the same kinds of microhabitats. Anolis trachyderma occurs in Amazonia and is superficially similar to A. nitens (also Amazonian) in ecology (Vitt and Zani 1996b; Vitt et al. 2002). Between 1994 and 1999, we assembled sufficient data to describe the ecology of A. fuscoauratus in considerable detail. We first present an overview of its ecology with all samples pooled. This approach provides a holistic view of its ecology over a large geographic area. We then consider geographic variation in its ecology by comparing data from six sites for which we have sufficient data. We explore possible reasons for differences among sites, including differences in the degree of phylogenetic similarity. Genetic similarity of A. fuscoauratus is high among sites (Glor et al. 2001). Finally, we comment on similarities between this anole in the Amazon rain forest and its closest relatives, and make some quantitative comparisons. We specifically comment on its similarities to A. limifrons in Central American rain forest. This comparison is particularly useful in that (i) A. fuscoauratus and A. limifrons are phylogenetically closely related but not each other s closest relative (see Glor et al. 2001) and (ii) A. limifrons appears quite similar morphologically and ecologically to A. fuscoauratus (e.g., Ballinger et al. 1970; Sexton et al. 1972; Talbot 1979). Methods and materials Study sites Anolis fuscoauratus was studied at six localities in Ecuador and Brazil (Fig. 1): (1) the Amazonian region of northeastern Ecuador in Sucumbíos Province (0 0, 76 10 W) near the Rio Cuyabeno (hereinafter Cuyabeno) during February April 1994 (rainy season); (2) terra firme Amazon lowland rain forest near the Rio Curuá-Una at Agropecuaria Treviso Limitada, approximately 101 km south and 18 km east of Santarém, Pará, Brazil (hereinafter Curuá-Una) (3 9 S, 54 50 W), during February April 1995 (rainy season); (3) approximately 5 km north of Porto Walter, Acre (hereinafter Rio Juruá) (8 15 S, 72 46 W), in undisturbed terra firme rain forest of the Juruá River basin during February April 1996 (rainy season); (4) the Rio Ituxi in the southwestern portion of Amazonas (hereinafter Rio Ituxi) (8 20 S, 65 43 W) in moderately disturbed rain forest during January April 1997 (rainy season); (5) approximately 40 km east of Guajará- Mirim, Rondônia, Brazil (hereinafter Rondônia) (10 19 S, 64 34 W), in tropical lowland forest during January March 1999 (rainy season); and (6) south of the Amazon River and nearly due south of Manaus, Amazonas (hereinafter Rio Solimões) (3 20 S, 59 4 W), in moderately disturbed rain forest during December January 1998 1999 (rainy season). Field protocol We conducted haphazard searches to locate lizards in forest (e.g., Vitt and Zani 1996c). Two to four researchers made observations virtually every day at each site. Because our field camps were either within forest or at forest edge, temporal bias in field data collection was minimal or nonexistent. We recorded the following data on each individual observed: habitat and microhabitat, time of day, whether it was sunny or cloudy, lizard s position with respect to sun (sun, shade, or filtered sun), distance above ground if on vegetation, and perch diameter. Eleven habitat categories were used for this analysis: (1) clearings, (2) low palm forest (flooded during heavy rains), (3) disturbed primary forest (some trees removed and some understory disturbed), (4) primary forest (terra firme, never flooded, no apparent disturbance), (5) secondary forest (forest that had clearly been cut in the past and had regenerated), (6) treefall (regardless of forest type), (7) stream (including stream edge and vegetation within streams), (8) swamp, (9) river forest (mixed forest that flooded during wet season), (10) forest edge (regardless of forest type), and (11) low

144 Can. J. Zool. Vol. 81, 2003 Fig. 1. Map of northern South America showing the locations of the Amazonian study sites. primary forest (hardwood forest that flooded during heavy rains). Microhabitats included (1) ground, (2) leaf (including palm leaves), (3) log, (4) leaf litter (including palm litter), (5) trunk, branch, or limb, (6) vine or tangle of vines, (7) exposed root (including buttress roots, stilt palm roots), (8) twig (any small stick), (9) shrub, and (10) stalks and fronds (bamboo and palm). Not all habitats and microhabitats occurred at every site, but those most commonly used by these lizards did (see Discussion). Additional habitat patches and microhabitats occurred at some sites but were not used by these lizards and therefore were not included. Cloacal (T b ), air (T a ), and substrate (T ss ) temperatures at the point of capture were measured for each lizard within 15 s of capture with Weber rapid-register thermometers (±0.2 C). Observations of active lizards took place between 0800 and 1800. Lizards were also collected at night while they were asleep. Laboratory protocol and analyses We returned lizards to our field laboratory within 2 h after collection, killed them by lethal injection of sodium pentabarbitol, and measured their snout vent length (SVL; to 1 mm). We also measured tail length and regenerated tail length, if any, to 1 mm, total mass to 0.1 g on a Acculab field balance, and head width, length, and height, body width and height, and hind leg and foreleg length to 0.1 mm with MaxCal electronic calipers. To compare morphology between the sexes, we first examined SVL of adults by restricting our analysis to lizards 40 mm SVL or greater. This size was arbitrarily chosen to exclude juveniles. SVL and body mass were compared between the sexes with Mann Whitney U tests. We performed analysis of covariance (ANCOVA) on log 10 -transformed morphological variables with log 10 SVL as the covariate to test for sexual dimorphism independent of body size. Log 10 transformations linearized the relationships. Stomachs were removed from lizards within 2 weeks after capture and either analyzed or placed in 70% ethanol for storage. When analyzed, stomachs were opened and prey items were separated and spread on a petri dish, identified to lowest possible taxonomic category (usually family), and measured for length and width. When removed from stomachs, prey items are compressed, with both legs and wings adhering to their body, making them easy to measure accurately. Our length and width measurements represent estimates of length and width of prolate spheroids. Width measurement took into account variation in insect shape. We grouped prey into 23 broad categories to summarize the diet: ants, blattarians, coleopterans, collembolans, dermapterans, dipterans, hemipterans, homopterans, hymenopterans (non-ant), insect larvae, isopods, lepidopterans, shed lizard skin, mantids, millipedes, mites, mollusks, neuropterans, orthopterans, pseudoscorpions, psocopterans, spiders, and termites. Volumes of individual prey were calculated with the formula for a prolate spheroid: V = 4/3π (length/2)(width/2) 2 using the program BugRun, a fourth-dimension-based analysis. The program produces dietary summaries, calculates mean prey size (length, width, and volume) for each lizard, estimates stomach volume based on total prey volume, and calculates niche breadth using the inverse of Simpson s (1949) measure (see Pianka 1973, 1986): 1 β= n 2 p i i= 1 where p is the proportional utilization of each prey type i. Niche-breadth values (β) vary from 1 (exclusive use of a single prey type) to n (even use of all prey). Data on prey size, stomach volume, and number of prey per stomach were log 10 -transformed for further analyses to normalize distributions. We used linear regression analysis to determine the influence of lizard body size on prey size and number of prey eaten. Log 10 SVL was the independent variable and prey data (log 10 -transformed) were the dependent variables. The effect of sex was determined using a ANCOVA with log 10 SVL as the covariate and sex as the class variable. We also plotted log 10 stomach volume with log 10 SVL to determine the relationship between stomach volume and lizard size and to estimate relative stomach fullness of lizards sampled. We make the assumption that a line parallel to the regression slope that intersects upper values approaches the relationship between stomach fullness and lizard size. Geographic comparisons We compared microhabitat use, body temperatures, activity periods, diets, and morphology among sites to determine whether intersite variation existed. For most such comparisons, standard statistical procedures were adopted. To compare diets among sites, we calculated a matrix of proportional utilization coefficients (p i ) and normalized electivities (e i ) using the combined diets of A. fuscoauratus from all sites as a measure of the total spectrum of prey eaten by the species. From these, we calculated geometric means of p i and e i (described in detail by Winemiller and Pianka 1990). Geometric means (g i ) balance bias associated with use of either p i or e i in further dietary analyses. We then calculated overlaps with the formula

Vitt et al. 145 n g ij ik i i φ jk = = 1 = 1 n n 2 2 gij gik i= 1 i= 1 n g where the symbols are the same as above but with j and k representing populations of A. fuscoauratus. An overlap matrix for all localities was calculated. To examine whether overlap values based on empirical data differed from what would be expected on the basis of a random sampling of the data, we performed two randomization (pseudocommunity) analyses. In the first, scrambled zeros, all values in the original consumer-resource matrix were randomized 1000 times and overlaps were calculated for each randomization. Number of resources and niche breadths remained intact. The second analysis, conserved zeros, was identical except that addresses in the consumer-resource matrix with zeros were not randomized with respect to position. Number of resources, niche breadths, and zero structure of the matrix were maintained. Resulting overlaps were ranked. Rank 1 indicates nearest neighbor and higher ranks represent correspondingly more distant neighbors in niche space. This allows statistical comparisons of overlaps resulting from the pseudocommunity analyses with measured overlaps at all nearest neighbor ranks. Statistical significance was based on the percentage of pseudocommunity overlaps falling below the actual community at each rank (Winemiller and Pianka 1990). Overlap values vary from 0 (no overlap) to 1 (identical diets). To assess morphological variation among sites, we first compared SVLs of all lizards 40 mm SVL or greater among sites, using a Kruskal Wallis test. Residuals were calculated from common regressions of each variable against SVL (all log 10 -transformed) to adjust morphological variables for lizard size and then entered into a principal components analysis (PCA) (see Miles 1994). Because large variation along axes was not attributed to geographic variation in some cases, we also included comparisons between the sexes to determine sources of variation. We compared the degree of lizard wariness among localities by developing an operational measure of detectability. First, because our daytime and nighttime searches were conducted identically across sites, we assume that variation among sites in our ability to detect lizards while they were active reflects variation in evasive behaviors of lizards rather than variation in our ability to detect them. We used sleeping lizards at night as a standard for relative abundance independent of wariness because no evasive behavior occurred unless lizards were physically disturbed. We make the assumption that differences in microhabitats used for sleeping were minimal compared with potential differences in behaviors while lizards were active and could avoid predators. Thus, our detectability index is Detectability = [number active/(number active + number asleep)] 100 High percentages indicate that lizards were easily detected while active, whereas low percentages indicate that lizards were not easily detected while active. Comparisons with closely related anoles We make comparisons with three closely related anoles for which we have similar data (A. humilis, A. limifrons, and A. trachyderma). Although considerable published data exist on ecology of A. limifrons and A. humilis (e.g., Pounds 1988), we restrict our analysis to data that two of us (L.J.V., P.A.Z.) collected in Nicaragua during 1993. Our protocol was nearly identical and as a result, data can be directly compared. Some of those data have appeared elsewhere (Vitt and Zani 1998; Vitt et al. 2002). To compare morphology among these anoles, we first compared body size between species. We then log 10 -transformed all morphological variables, calculated residuals from the common regressions of all morphological variables with SVL to produce relatively size-adjusted variables, and entered the residuals in a PCA. Voucher specimens of all lizard species used in this study were preserved in 10% formalin and later transferred to 70% ethanol for permanent storage, and are housed in the herpetology collections of the Museu P. Emílio Goeldi in Belém, Brazil, the herpetology collections at the Instituto de Pesquisas da Amazonia (INPA) in Manaus, Brazil, the Museo de Zoología de la Pontificia Universidad Católica (QCAZ) in Quito, Ecuador, and the Sam Noble Oklahoma Museum of Natural History in Norman. Statistics were performed with StatView 5.0 (SAS Institute Inc. 1998) unless indicated otherwise. Means appear ±1 standard error. Results Anolis fuscoauratus in the Amazon Habitat and microhabitat use Nearly half (41.7%) of all active A. fuscoauratus were in undisturbed primary forest (Fig. 2). Many lizards were also observed or captured in other relatively undisturbed forest, including low primary forest (13.9%) and river forest (8.0%), the latter habitat occurring only at the Rio Ituxí site. Substantial numbers (13.6%) were found at the edges of natural treefalls as well. The most common microhabitats used by A. fuscoauratus were trunks, branches, and limbs of trees (39.1%), leaf litter (15.6%), and leaves on trees and shrubs (11.8%), but logs, vines, and twigs were frequently used as well (Fig. 3). Perch height was 0.98 ± 0.05 m (n = 308) and perch diameter (excluding leaves) was 12.85 ± 2.26 cm (n = 251). One hundred and eight sleeping A. fuscoauratus were found by searching all forest microhabitats at night with headlights. Seventy (64.8%) were in primary forest, 11 (10.2%) were in river forest, 6 (5.6%) were in forest edge, 6 (5.6%) were in low primary forest, 5 (4.6%) were in secondary forest, 4 (3.7%) were in swamps, 3 (2.8%) were in disturbed primary forest, 2 (1.9%) were in treefalls, and 1 (0.9%) was in a clearing. Fifty-five (50.9%) were at the ends of branches of trees or shrubs, 39 (36.1%) were on leaves of plants, 11 (10.2%) were on the end of twigs, 2 (1.9%) were in leaf litter, and 1 (0.9%) was on a vine. For those asleep on branches, twigs, or vines, perch diameter was 0.37 ± 0.12 cm. Sleeping anoles were 2.11 ± 0.20 m off ground (n = 107). Sleeping perches (not including leaves) were 0.96 ± 0.27 cm in diameter (n = 74). Lizards were always perched with the head toward the end of the branch. Leaves with sleeping lizards were 17.4 ±

146 Can. J. Zool. Vol. 81, 2003 Fig. 2. Habitat categories in which Amazonian Anolis fuscoauratus were observed, ranked from least to most frequently used habitat. Fig. 4. Daily activity cycle for Amazonian A. fuscoauratus (open bars), showing the percentage of lizards that were active in sun ( ). Fig. 5. Relationship between body temperature (T b ) and substrate temperature (T ss ) at exact point of capture for A. fuscoauratus (T b = 0.817(T ss ) + 6.531). Fig. 3. Microhabitat categories in which Amazonian A. fuscoauratus were observed, ranked from least to most frequently used microhabitat. 7.4 cm in length by 7.4 ± 0.92 cm in width. When perches were physically disturbed, sleeping lizards dropped to the ground, usually remaining motionless. Activity and body temperatures Activity occurred from just after sunrise until just before sunset (Fig. 4). Once the sun filtered through the forest in the morning, individuals became active, basked for a short time in sun, and remained active, usually in more shaded microhabitats, for the entire day. Slightly more lizards were seen when it was cloudy (223, 59.9%) than when the sun was shining (149, 40.1%). For 368 individuals observed active, 224 (60.9%) were in shade, 92 (25.0%) were in filtered sun, and 52 (14.1%) were in sun. T b of 81 active lizards was 28.7 ± 0.2 C. The lowest T b recorded for an active lizard was 25.7 C and the highest was 33.8 C. T ss associated with these lizards was 27.1 ± 0.2 C and T a was 27.0 ± 0.1 C. T b was significantly higher than T ss and T a (F [1,160] = 44.4, P < 0.001, and F [1,160] = 56.2, P < 0.001, respectively). T b was 1.6 ± 0.1 C higher than T ss and 1.7 ± 0.1 C higher than T a. Neither sun availability (F [2,73] = 2.3, P = 0.13) nor exposure (ANOVA, F [2,73] = 2.0, P = 0.14) influenced lizard T b and the interaction term was nonsignificant (ANOVA, F [2,73] = 1.2, P = 0.32). Overall, lizard T b was significantly correlated with T ss (Fig. 5; R 2 = 0.54, F [1,79] = 93.6, P = 0.0056). We pooled T b s for 2-h periods beginning at 08:00 to examine T b variation during the day (Fig. 6). Lizard T b varied significantly through the day (ANOVA, F [4,76] = 5.8, P = 0.0004). T b s were lower during the period 1000 1159 than during the periods 1200 1359 and 1400 1559 (P < 0.05, Games Howell post-hoc test; see Games and Howell 1976). When the effect of T ss on T b was adjusted by calculating regression residuals and the ANOVA was repeated, the differences disappeared (F [4,76] = 1.7, P = 0.1515). Thus, even though T ss influences T b, lizards maintained T b slightly higher than ambient temperature. Diet When data from all localities were included, spiders, beetles, termites, ants, and grasshoppers were most common in the diet numerically (Table 1). Volumetrically, grasshoppers,

Vitt et al. 147 Table 1. Diet of Anolis fuscoauratus (all localities pooled). Prey category No. % No. Vol. (mm 3 ) % vol. Frequency Collembolans 7 1.54 0.99 0.02 3 Orthopterans 37 8.11 1115.49 20.62 33 Mantids 1 0.22 73.18 1.35 1 Blattarians 10 2.19 443.44 8.2 10 Hemipterans 4 0.88 31.34 0.58 4 Homopterans 30 6.58 574.71 10.62 23 Coleopterans 57 12.5 238.29 4.41 35 Termites 52 11.4 319.31 5.9 11 Lepidopterans 3 0.66 94.85 1.75 3 Neuropterans 1 0.22 49.02 0.91 1 Dermapterans 5 1.1 16.97 0.31 3 Psocopterans 1 0.22 0.37 0.01 1 Dipterans 13 2.85 42.28 0.78 13 Hymenopterans (non-ant) 13 2.85 54.53 1.01 11 Ants 51 11.18 481.29 8.9 32 Insect larvae 42 9.21 560.11 10.35 33 Spiders 76 16.67 906.5 16.76 64 Mites 20 4.39 0.55 0.01 3 Pseudoscorpions 2 0.44 1.32 0.02 2 Millipedes 18 3.95 218.27 4.03 17 Isopods 8 1.75 119.04 2.2 7 Mollusks 1 0.22 6.4 0.12 1 Lizard shed skin 4 0.88 61.17 1.13 4 Sum 456 100 5409.42 100 Niche breadth 10.55 8.64 Note: Frequency is the number of lizards eating prey in a particular category. Fig. 6. Daily progression of A. fuscoauratus body temperature (T b ) and substrate (T ss ) and air (T a ) temperatures. Fig. 7. Relationship between body size (snout vent length) and total stomach contents for A. fuscoauratus. Variations above and below the regression serve as relative measures of stomach fullness. spiders, homopterans, and insect larvae dominated the diet. Four hundred and fifty-six prey items were 5.13 ± 0.15 mm (range 0.41 22.58 mm) in length, 1.53 ± 0.04 mm (0.24 5.92 mm) in width, and 11.86 ± 0.96 mm 3 (0.01 201.94 mm 3 ) in volume. The relationship between total volume of stomach contents (an estimate of stomach size) and lizard size (SVL) was significant (R 2 = 0.522, F [1,162] = 178.8, P < 0.0001), but many of these anoles did not have a full stomach when sampled (Fig. 7). Although mean prey volume was highly correlated with lizard SVL, head width, and head length, the greatest portion of the variation in mean prey volume was attributable to variation in head width (R 2 = 0.650, F [1,162] = 303.5, P < 0.0001; Fig. 8). Intact prey were found in 165 of 217 stomachs (76%). Among lizards containing prey, the number of prey per stomach was 5.201 ± 0.046 (4 8) and was significantly associated with lizard body size (R 2 = 0.354, F [1,162] = 90.5, P < 0.0001; Fig. 9). Larger lizards ate larger prey and more of them.

148 Can. J. Zool. Vol. 81, 2003 Fig. 8. Relationship of mean prey size (volume) to lizard head width for A. fuscoauratus, showing that larger lizards generally eat larger prey than smaller lizards do. Fig. 10. Size (snout vent length) distribution of A. fuscoauratus used in this study, showing that females tend to reach a larger size than males. 0.6726; intercepts, F [1,151] = 0.17, P = 0.6808). The only apparent sexual dimorphism is in size (SVL) and the presence of a dewlap in males. Fig. 9. Relationship of number of prey per stomach to lizard size (snout vent length) for A. fuscoauratus, showing that larger lizards generally eat more prey than smaller lizards do. Body size and sexual dimorphism The smallest A. fuscoauratus captured was a hatchling male of 16.9 mm SVL. The largest was a female of 52 mm SVL (Fig. 10). When only lizards of 40 mm SVL or greater were considered, females were 45.8 ± 0.2 mm SVL (40 52) and weighed 1.7 ± 0.03 g (1.1 2.3 g, n = 108) and males were 43.7 ± 0.3 (40 48 mm, n = 47) and weighed 1.3 ± 0.03 g (0.9 1.7 g, n = 47). Females were significantly larger than males in both SVL (Mann Whitney U test, Z = 5.14, P < 0.0001) and mass (Mann Whitney U test, Z = 8.24, P < 0.0001). Males and females were similar in relative head width (ANCOVA, slopes, F [1,151] = 0.93, P = 0.3354; intercepts, F [1,151] = 0.88, P = 0.3491) and relative head length (ANCOVA, slopes, F [1,151] = 0.66, P = 0.4191; intercepts, F [1,151] = 0.65, P = 0.4223). Relative hind-limb and forelimb lengths were also similar (hind limb: ANCOVA, slopes, F [1,151] = 0.00001, P = 0.9972; intercepts, F [1,151] = 0.001, P = 0.9812; forelimb: ANCOVA, slopes, F [1,151] = 0.18, P = Geographic comparisons Body temperatures To compare lizard body temperatures among sites, we first eliminated data from the Curuá-Una site because we had only a single data point. We then performed a ANCOVA with T b as the dependent variable, T ss as the covariate, and locality as the class variable. No effect of locality was detectable when variation in T b associated with T ss was accounted for (F [4,70] = 1.6, P = 0.1872) and the interaction term was not significant (F [4,70] = 1.7, P = 0.1659). Variation in T b among sites resulted from the effect of variation in T ss on T b. Diets Considerable variation in diets was apparent among sites (Fig. 11). Lizards at Cuyabeno, Curuá-Una, Rondônia, and Rio Solimões ate a greater diversity of prey than those at Rio Juruá and Rio Ituxí. At Rio Juruá and Rio Ituxí small grasshoppers and crickets dominated the diet, exceeding 45% volumetrically. Spiders were much more common at Cuyabeno and Curuá-Una, but beetles, termites, and lepidopterans composed much of the remaining diet at Curuá-Una, whereas homopterans, beetles, and insect larvae contributed more at Cuyabeno. Dietary overlaps varied from relatively high (0.720) between the Rio Ituxi and Rio Juruá populations to very low between the Curuá-Una and Rio Ituxi populations (0.174; Table 2). Pseudocommunity analysis comparing diets of lizards among sites revealed nonrandom use of prey among sites. Measured dietary overlaps were significantly higher than those based on chance at all ranks when all dietary data were randomized but were not different when zero positions in the consumer-resource matrix remained intact (Fig. 12). Variation in measured overlaps was also much lower than that in randomizations. At all but two sites (Rio Juruá and Rio Ituxí), A. fuscoauratus were generalists with respect to each other even though they ate different sets of prey (Fig. 13). Rio Juruá and Rio Ituxí lizards ate mostly orthopterans. Overlaps as a function of nearest neighbor rank dropped relatively slowly, indicating that in most comparisons, differences among populations remained uniform. The greatest exception was the Rio Ituxi population, which dropped from an overlap exceeding 0.7 with its closest

Vitt et al. 149 Fig. 11. Comparisons of prey use by A. fuscoauratus among six Amazonian study sites. in size-adjusted morphology of adult A. fuscoauratus (Table 3, Fig. 16). Four principal- component axes accounted for 81.8% of the variation in size-adjusted morphology. The first axis, accounting for 42.4% of the variation, describes a gradient based on mass, body width, and body height, all measures of relative robustness. No effect of locality was detected (F [5,190] = 2.0, P < 0.0818) but a significant effect of sex was apparent (F [5,190] = 96.3, P < 0.0001). Females were heavier bodied than males. The second axis, accounting for 17.3% of the variation, describes a gradient based on relative hind-leg and foreleg lengths. A significant effect of locality exists (F [5,190] = 10.7, P < 0.0001), with Acre lizards having shorter limbs than Cuyabeno and Rondônia lizards, Curuá-Una lizards having shorter limbs than Rondônia lizards, and Cuyabeno lizards having longer limbs than Curuá-Una lizards (P < 0.05, Games Howell post-hoc test). No effect of sex on limb length was detected (F [5,190] = 0.5, P = 0.5031). The third axis, accounting for 11.8% of the variation, describes a gradient based on head height and head width. Similar to relative limb length, a significant effect of locality exists (F [5,190] = 3.5, P < 0.0052). Sex did not affect head width and depth (F [5,190] = 2.0, P = 0.0848). The fourth axis, accounting for 10.3% of the variation, describes a gradient based on head length. No effect of locality (F [5,190] = 1.0, P < 0.4065) or sex (F [5,190] = 1.5, P = 0.2205) was apparent. Detectability while active Detectability indices varied from 6.12% at the Curuá-Una site to 93.79% at the Rio Juruá site (Table 4). While active, lizards at the Curuá-Una site were so wary that we rarely found them even though our nighttime searches indicated that they were among the most common lizards. At Rio Juruá, A. fuscoauratus was highly visible while active. neighbor, the Rio Juruá population, to less than 0.4 compared with its next closest neighbor, the Rondônia population. Size of prey also varied among sites (Fig. 14; F [5,450] = 10.1, P < 0.0001). Anolis fuscoauratus from Rio Solimões ate significantly smaller prey than those from all other sites (Games Howell post-hoc test, all P < 0.05). Morphology and sexual dimorphism Mean SVL (for lizards 40 mm SVL or greater) varied significantly (Fig. 15; ANOVA, F [5,191] = 7.0, P < 0.0001) among sites. Rio Juruá anoles were smaller than Rio Solimões anoles, and Rio Solimões and Cuyabeno anoles were smaller than those from Rio Ituxi and Rondônia (Games Howell posthoc test, all P < 0.05). Variation among sites was also evident Comparisons with other anoles Significant body-size variation exists among the four anole species, A. trachyderma being the largest and A. humilis the smallest (Table 5). Considerable morphological variation exists as well. The PCA describes three size-adjusted axes that account for more than 75% of Bauplan variation among these four species (Table 6, Fig. 17). Factor I describes a gradient based on overall robustness weighted by relative mass, body height, body width, and, to a much lesser degree, head width. Factor II describes a gradient based on relative hind-limb and forelimb length. Factor III describes a gradient based on relative head length. Overall, A. humilis is the most robust species and A. limifrons is the most streamlined. Anolis fuscoauratus and A. limifrons have nearly identical Baupläne. Anolis trachyderma has the longest limbs, whereas A. fuscoauratus has the shortest. Anolis fuscoauratus is not most similar to A. limifrons in limb length; instead, it is most similar to A. humilis. Anolis humilis has the widest head, A. trachyderma the narrowest. Discussion Geographic ecology Anolis fuscoauratus is a small forest anole that lives off the ground on vines, shrubs, and tree limbs in the Amazon rain forest. It tends to be most common in undisturbed forest and rare in open habitat patches, although it is often found at

150 Can. J. Zool. Vol. 81, 2003 Table 2. Dietary overlaps based on volumetric data for A. fuscoauratus. Amazonas Curuá-Una Cuyabeno Rio Ituxi Rio Juruá Curuá-Una 0.563 Cuyabeno 0.523 0.446 Rio Ituxi 0.301 0.174 0.240 Rio Juruá 0.440 0.319 0.358 0.720 Rondônia 0.536 0.309 0.570 0.353 0.408 Fig. 12. Results of pseudocommunity analysis comparing diets of A. fuscoauratus among study sites. Overlaps from empirical results are significantly higher and less variable than those of type 1 simulations, indicating that diets are more similar among sites than would be expected by chance. Fig. 13. Results of pseudocommunity analysis of diets of A. fuscoauratus portrayed to show changes in overlaps among sites as a function of nearest neighbor. Steep slopes indicate progression toward dietary specialization relative to other populations. For example, the diet of Rio Ituxi lizards is very similar to that of Rio Juruá lizards but very different from the diets of those at other sites. Points have been offset for clarity. At rank 1, points for nearest neighbors would overlap; at other ranks, one population s nearest neighbor might not be the other s. the edge of treefalls. Most lizards observed are first found in shade or filtered sun, but basking in direct sun does occur. Activity occurs throughout the day whether it is sunny or cloudy, and lizards are most likely to be observed basking in sun early in the morning. T b of active individuals is significantly correlated with both T ss and T a but is slightly higher (1 2 C throughout the entire day) than both, indicating that the lizards gain some heat by short periods of exposure to either direct sun or filtered sun. This appears to be common among Amazon lizards in forested habitats. For example, A. nitens, a forest-floor species, also maintains T b slightly higher than T ss and T a even though it is usually found in shade (Vitt and Zani 1996b), as do the gymnophthalmids Neusticurus ecpleopus (Vitt et al. 1998a) and Prionodactylus eigenmanni (Vitt et al. 1998b). Similar patterns have been observed in Costa Rican and North American anoles (e.g., Clark (1973) and Clark and Kroll (1974), respectively). Although correlations between T b and both T ss and T a in rain-forest lizards suggest that they are thermal conformers, the observation that the average T b is higher than either indicates that the lizards do not simply conform to the temperatures of their immediate microhabitats. It is possible that all ectotherms living in the forest have similar problems gaining heat, so that maintaining T b slightly higher than surrounding microhabitats provides small lizards with a behavioral advantage over larger bodied ectothermic predators. When all else is equal, larger bodied predators would gain heat at a slower rate and thus lag in T b behind small lizards like A. fuscoauratus as the day progresses. They would also retain heat longer at the end of the activity period. Because the performance of diurnal ectothermic vertebrates is usually tied to temperature (Hertz et al. 1983, 1988; Garland 1994), slightly warmer A. fuscoauratus might be more likely to escape ectothermic predators than cooler ones. The frequency of defensive responses is known to in-

Vitt et al. 151 Fig. 14. Prey-size distributions for A. fuscoauratus from six Amazonian localities. Fig. 15. Body-size comparisons of A. fuscoauratus from six Amazonian localities. crease with temperature in lizards (Hertz et al. 1982). Moreover, maintaining higher T b increases lizards abilities to capture, feed on, and digest prey (Van Damme et al. 1991) and to detect predators on the basis of chemical cues (Van Damme et al. 1990). Even though A. fuscoauratus maintains T b slightly higher than its surroundings, its T b while active is low compared with that of heliothermic Amazon lizards such as Kentropyx spp. (Vitt 1991b; Vitt and Carvalho 1992; Vitt et al. 1995, 1997b), Cnemidophorus (Vitt et al. 1997a), Ameiva spp. (Vitt and Colli 1994; Sartorius et al. 1999), and Mabuya nigropunctata (Vitt et al. 1997b). It remains unknown whether the performance of A. fuscoauratus and other rain-forest lizards is higher than expected based on T b alone as in some nocturnal gekkonids (e.g., Autumn et al. 1994, 1999). The impact of relatively low rain-forest temperatures

152 Can. J. Zool. Vol. 81, 2003 Table 3. Oblique solution reference structure from principal-components analysis (PCA) of sizeadjusted morphological variables for A. fuscoauratus from six Amazonian localities. Variable Factor I Factor II* Factor III* Factor IV Mass 0.844 0.071 0.180 0.004 Head width 0.309 0.081 0.732 0.019 Head length 0.063 0.014 0.002 0.985 Head height 0.056 0.004 0.932 0.013 Body width 0.897 0.0001 0.010 0.003 Body height 0.806 0.100 0.018 0.069 Hind-leg length 0.136 0.914 0.100 0.024 Foreleg length 0.004 0.804 0.168 0.028 Eigenvalue 3.392 1.384 0.945 0.825 Percentage of variation 42.4 17.3 11.8 10.3 Note: The analysis included only lizards of 40 mm SVL or greater. *Significant difference between localities. Fig. 16. Plots of factors II and III of a PCA comparing sizeadjusted morphology of A. fuscoauratus from six Amazonian localities. Factor I is not shown because no effect of locality was detected. on T b of A. fuscoauratus may account for its extended period of activity. Populations of the desert lizard Sceloporus merriami maintain longer activity periods in habitats with relatively lower daily temperatures than they do in habitats with higher temperatures, suggesting that temperature affects daily activity (Grant 1990). Activity off the ground as well as in shaded forest offers some protection for A. fuscoauratus from heliothermic lizards that often prey on other lizards. The diet of A. fuscoauratus consists of a wide variety of arthropods with additional prey, including mollusks. Like many other anoles, they also eat their own shed skin. The most common prey volumetrically were taxa with relatively low proportions of exoskeleton: orthopterans, homopterans, insect larvae, and spiders. Whether these lizards select such prey or the frequency with which they are eaten simply reflects differential availability remains undetermined. Because most iguanians detect prey visually and have poorly developed chemosensory systems (Cooper 1995, 1997; Schwenk Table 4. Anolis fuscoauratus observed active during the day compared with those observed asleep at night at six Amazonian sites. Locality No. active No. asleep Total observed Detectability (%) Cuyabeno 65 19 84 77.38 Curuá-Una 3 46 49 6.12 Rio Juruá 166 11 177 93.79 Rio Ituxí 13 7 20 65.00 Rondônia 91 17 108 84.26 Rio Solomões 36 8 44 81.82 Note: High detectability values indicate that lizards were relatively easy to observe while active. 2000), prey discrimination based on chemical cues probably does not occur. Because ants are generally highly active and abundant, these lizards may selectively avoid ants on the basis of visual cues; alternatively, they may select more profitable prey. Based on dietary analyses of other lizard species from most of these localities (e.g., Vitt and Zani 1996a, 1996b, 1996c; Vitt et al. 1999), ants are available. Capturing food does not appear to be difficult for these lizards because most individuals, regardless of locality, had a relatively full stomach. Prey may be available in high enough numbers that the lizards can afford to pass by low-energy prey items, thus accounting for the low proportion of ants in the diet volumetrically. In contrast, some other Amazonian anoles eat substantial proportions of ants (e.g., A. nitens; Vitt et al. 2001). Variation across sites At all sites, A. fuscoauratus was arboreal and usually found in undisturbed forest. Forest structure varied somewhat among sites (e.g., Rio Ituxí had extensive river forest, whereas other sites had mostly primary forest), but the lizards used whatever forest was available. Similarly, A. fuscoauratus at all sites were most frequently found on elevated perches. We conclude that across their Amazonian range, these lizards use arboreal perches in forest that offer shade. At all sites, activity occurred throughout the day. At the wettest and coolest site (Cuyabeno), more were seen about midday that at other times, whereas at the warmest and driest site (Rondônia), activity was slightly bimodal (morning and late afternoon). Thus, activity appears to be impacted to

Vitt et al. 153 Table 5. Comparisons of morphological and ecological characteristics of four closely related Anolis species. A. fuscoauratus A. humilis a A. limifrons a A. trachyderma b Maximum SVL (mm) Males 48 40 39 56 Females 52 37 44 61 Maximum mass (g) Males 1.7 1.3 1.1 2.8 Females 2.3 1.4 1.6 4.6 T b ( C) c 28.7 ± 0.2 27.8 ± 0.6 29.1 ± 0.3 27.8 ± 0.2 Microhabitat d TBL LL, TBL TBL LL, TBL Perch height (m) c 0.98 ± 0.05 0.7 ± 0.09 0.35 ± 0.05 0.39 ± 0.04 Perch diameter (cm) c 12.85 ± 2.26 43.0 ± 41 3.2 ± 0.6 10.4 ± 1.8 Diet e O, S, HO, IL P, S, IL HO, O, IS, IL O, MO, IL, S Mean prey volume (mm 3 ) c 11.86 ± 0.96 18.2 ± 5.9 16.8 ± 5.0 23.9 ± 4.3 No. of prey per stomach c 5.201 ± 0.046 5.6 ± 1.9 3.3 ± 0.4 2.1 ± 0.2 a Data from Vitt and Zani (1998) and L.J. Vitt and P.A. Zani, unpublished data. b Data from Vitt et al. (2002) and unpublished data. c Values are given as the mean ± SE. d Categories are as follows: TBL, trunk, branch, limb; LL, leaf litter. e Categories are as follows: O, orthopterans; S, spiders; IL, insect larvae; HO, homopterans; IS, isopods; MO, mollusks. The order of prey categories corresponds to rank by volume. Table 6. Oblique solution reference structure from a PCA of size-adjusted morphological variables of four closely related anole species. Fig. 17. Plots of factors I and II of a PCA comparing sizeadjusted morphology of A. fuscoauratus with that of its four closest relatives (phylogeny is based on Glor et al. 2001). Variable Factor I Factor II* Factor III* Mass 0.809 0.237 0.051 Head width 0.585 0.460 0.231 Head length 0.001 0.199 0.850 Head height 0.324 0.409 0.437 Body width 0.745 0.308 0.014 Body height 0.778 0.137 0.034 Hind-leg length 0.001 0.886 0.125 Foreleg length 0.022 0.895 0.035 Eigenvalue 3.266 2.173 0.748 Percentage of variation 40.0 27.2 9.3 *Significant difference between species based on ANOVA; all P < 0.0001. some degree by temperature, moisture, or a combination of the two. T b varied among sites. Lizards at Cuyabeno had a significantly lower T b than Rondônia lizards. Even when the effect of T ss on T b was removed, Rondônia lizards had higher T b than Cuyabeno lizards, suggesting that some physiological adjustment to a history of differences in thermal environment may have occurred. Diets varied geographically but lizard diets at some sites were more similar to each other than expected by chance when zeros in the consumer-resource matrix were not randomized (type 2 randomization). This can be attributed largely to the high similarity in diets between Rio Ituxí and Rio Juruá lizards, particularly their use of orthopterans. Interactions with sympatric Anolis species might influence the diets of A. fuscoauratus at the local level, but this is not apparent from survey data for anoles (Table 7). The Rio Juruá and Rio Ituxí populations are similar in that they do not occur with A. trachyderma, a similar-sized and closely related anole that lives at the leaf litter bush interface (Vitt et al. 2002). However, A. trachyderma also does not occur at the Rondônia and Rio Solimões sites, where A. fuscoauratus diets are quite different from those at the Rio Juruá and Rio Ituxí. The total number of lizard species varies among localities as well (Table 7), but no effect of number of sympatric lizard species on the ecology of A. fuscoauratus is evident. Alternatively, among-population variation in diets may reflect differences in relative availability of major prey types (e.g., orthopterans, spiders, insect larvae, and homopterans) among sites. Additional work will be necessary to determine the causes of geographic variation in the diets of these lizards. Mean SVL of adults varied geographically as did some aspects of size-adjusted morphology. The underlying cause

154 Can. J. Zool. Vol. 81, 2003 Table 7. Anole species occurring with A. fuscoauratus at six Amazonian sites. Locality A. nitens A. trachyderma A. ortonii A. punctatus A. transversalis No. of lizard species a Cuyabeno + + + + 23 Curuá-Una + + + 25 Rio Juruá + _ + + 29 Rio Ituxí + + + + 30 Rondônia + + + + 24 Rio Solomões + + + 19 Note: + indicates presence and indicates absence. a Total lizard species diversity for each locality (L.J. Vitt, T.C.S. Avila-Pires, P.A. Zani, and S.S. Sartorius, unpublished data). of morphological variation in these anoles remains unknown. Ecologically, the populations are similar, at least with respect to their predominantly arboreal habits. Dietary differences do occur, but whether such differences could result in population differentiation remains unknown. However, A. fuscoauratus populations that have been studied vary little phylogenetically (Glor et al. 2001). Comparisons with close relatives As shown in Table 5, the ecological attributes of anoles in the A. fuscoauratus A. humilis A. limifrons A. trachyderma clade are similar, particularly when they are considered as sympatric species pairs. Anolis fuscoauratus A. trachyderma in Amazonia are similar to A. limifrons A. humilis in Central America in that the first of each pair is more arboreal, has a higher T b, and tends to use higher perches. Both Amazonian anoles are larger in body size than either Central American species. However, in the Amazon, the more arboreal of the species pair (A. fuscoauratus) is smaller, whereas in Central America, the more arboreal of the species pair (A. limifrons) is the larger. In view of the observation that A. fuscoauratus (Amazonia) and A. humilis (Central America) are each other s closest relatives and A. trachyderma (Amazonia) and A. limifrons (Central America) are each other s closest relatives (Glor et al. 2001), convergence likely has occurred in the evolution of ecomorphs between Amazonia and Central America, similar to that found in island anoles (Losos 1992, 1994, 1995; Losos et al. 1998; Williams 1969, 1972, 1983). Nevertheless, ecomorphs of mainland anoles show striking differences when compared with island anoles in the context of entire faunas (Irschick et al. 1997). Conclusions Many autecological studies combine data from all study sites and thus present a broad perspective on a particular lizard species (e.g., Vitt 1991a, 1991b; Vitt and Blackburn 1991). Such studies provide good overviews of the range of niche characteristics (food, place, and time; Pianka 1973) of a species. Other studies specifically address ecological variation on a geographic level (e.g., Pianka 1970; Blasco and Romero 1985; James 1994). Such studies demonstrate that ecological traits of lizard species vary to some degree geographically. Nevertheless, lizard species generally occur in the same microhabitats, are active at about the same time of day, and eat the same kinds of things regardless of where they occur. One factor that might account for the ecological similarity of A. fuscoauratus among sites is relative relatedness. Anolis fuscoauratus from these localities exhibit very little detectable molecular evolution compared with other species of anoles from the same sites (Glor et al. 2001). Ecological similarity among sites may simply reflect phylogenetic similarity. Because of its arboreal habits, A. fuscoauratus likely has maintained genetic consistency across great distances as a result of nearly continual gene flow. Unlike terrestrial anoles, such as A. nitens, A. fuscoauratus moves about in vegetation, on vines, and in trees. Consequently, it is presumably less affected by potential dispersal barriers such as rivers or flooded forest. A different explanation for variation in body size and ecology exists for the primarily terrestrial Amazonian anole A. nitens. It apparently has experienced a long history of isolated populations (see Glor et al. 2001). Although variation occurs among populations of A. fuscoauratus, in a general way its ecology is similar across the Amazon. It occurs in shaded forest, uses branches, limbs, and trunks of trees as well as vines, maintains relatively low T b, and avoids microhabitats exposed to direct sun during most of the day. Like most forest lizard species, A. fuscoauratus is at risk of significant population declines if forest is modified in such a way that both the structural and thermal environments important for its survival are compromised. Acknowledgments First, we thank all of the people who either helped coordinate our research in one or more areas or helped us collect data: W.E. Magnusson, A. Lima, M.C. Araújo, J.P. Caldwell, V. Oliveira, Marcelo Scheffer, Miguel Schultz, and Antonio Leite. Brazilian agencies contributing to logistics or research and collecting permits include SOS Amazônia and the Acre Union of Seringueiros in Rio Branco, the CEMEX Timber Company in Pará, INPA, the CNPq (Portaria MCT No. 170, de 28/09/94), the Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis (permit No. 073/94- DIFAS), and the Museu Paraense Emílio Goeldi in Belém. Brazilian research was conducted under a research convenio between the Sam Noble Oklahoma Museum of Natural History and the Museu Paraense Emílio Goeldi. We thank L. Coloma for coordinating our Amazonian research in Ecuador and the QCAZ in Quito for logistic support. The Estación Biológica de Cuyabeno in Ecuador was made available by the Universidad Católica and permits were issued by Ministry of Agriculture and Livestock of the Republic of Ecuador. All animals were treated in accordance with federal, state, and university regulations (Animal Care Assurance 73- R-100, approved 8 November 1994). We thank American Airlines and Varig Airlines for allowing excess baggage. National Science Foundation grants DEB-9200779 and DEB-