We sought to clarify relationships among morphometrics, behavior, and ecological variables for 21 species of raptors in the Philippines. Morphological space was defined by 42 external characters analyzed as shape variables with respect to body length (without tail). Seventeen variables were used to characterize habitat and five to characterize foraging behavior. Three PCA components accounted for 68% of the total variance in the habitat data and separated species living in dense forests from those using degraded habitats or coastal areas. Two PCA components explained 81% of the variance in hunting mode, with transitions from sit-and-wait to flap-gliding, and with contrasts between soaring and flap-gliding. Three PCA components accounted for 70.5% of the variance in morphological shape. The first component separated species with narrow wings and less-pronounced notches from species with broad, deeply notched wings. The second and third components were associated with the contrast between pointed and rounded wings and prey-capture apparatus (feet, bill). Five morphological characters were highly correlated (R = 0.934) with the first principal component of the habitat data, indicating that species inhabiting forested habitats have square tails, rounded wings, and strong claws. Hunting mode and habitat also were closely related (R = 0.900). Soaring correlated well with the number of notched primaries, tail shape, and measures of the trophic apparatus, but poorly with wing loading and aspect ratio. Behavior, ecology, and morphology of this subset of raptors were closely interrelated. Among the Philippine raptors, species that inhabit rain forests are the most endangered, and we suggest that morphological constraints limit their use of secondary habitats. Received 21 January 1997, accepted 2 February 1998.

Konrad Lorenz-Institute for Comparative Ethology, Austrian Academy of Sciences, A-1160 Vienna, Austria DIFFERENCES IN EXTERNAL MORPHOLOGY among species can reliably predict differences in habitat use and foraging behavior That this is true is the premise of ecomorphological anal- yses of bird communities (Karr and James 1975, Leisler 1980, Miles and Ricklefs 1984, Niemi 1985, Miles et al. 1987, Wiens 1989). Yet, this as- sumption is rarely tested with data on behavior and habitat acquired simultaneously in the field, and comparatively few studies have at- tempted to clarify relationships among mor- phometry and the behavior or environment of the species in question (but see Leisler et al. 1989, Price 1991, Landmann and Winding 1993, Stiles 1995). Such work is needed to corroborate premises of ecomorphology and to provide the necessary link between resource use, perfor- mance, and morphology demanded for a deep- er understanding of the relationship between Present address: Museum of Natural History, De- partment of Vertebrate Zoology, Bird Collection, Burgring 7, A-1014 Vienna, Austria. 2 Address correspondence to this author. E-mail: h.winkler@klivv. oeaw. ac.at the design of birds and their environment (Led- erer 1984, Winkler 1989, Wainwright 1991). Common difficulties in ecomorphological anal- yses of this kind are character sets that contain only few or irrelevant characters, or both (Leis- ler and Winkler 1997). Analyses of closely re- lated species are usually most successful in showing ecomorphological relationships (Led- erer 1984, Leisler and Winkler 1985), whereas hypotheses of wider phylogenetic scope face increasing difficulties because of oversimplifi- cation (Lederer 1984) and because phylogeny may confound analyses at the community level (Miles et al. 1987). To strike a balance between the difficulties noted above, we studied the diurnal birds of prey of the Philippines. This group is ecologi- cally and phylogenetically well defined but is heterogeneous enough to mediate between studies confined to closely related species and those of guilds and communities. So far, no in- tegrative investigations are available for birds of prey, although some studies on size differ- ences (Schoener 1984) and morphology (Ro- chon-Duvigneaud 1952, Voous 1969, Nieboer 1973, Brown 1976, Bierregaard 1978, Jaksi 1985, Kemp and Crowe 1994, Hertel 1995, Jen- kins 1995) exist in which morphological pat- terns were related to some aspects of ecology. However, studies on raptors that combine data on hunting modes and habitat use have not been attempted. The Philippines is an excellent area for this kind of investigation because a large variety of raptor species can be observed in an area of less than 300,000 km 2. Moreover, habitats in the Philippines span distinctive eco- logical gradients from tropical primary rain forest to deforested regions and coastlines. Originally, the Philippines were almost com- pletely covered by tropical rain forest. Because the proportion of forested area was reduced from 58 to 17% between 1932 and 1991, several species of raptors currently are threatened by extinction (Collins et al. 1991, Dickinson et al. 1991, ICBP 1992, Collar et al. 1994). Our objectives were to: (1) examine the strength, consistency, and functional basis of the relationships among foraging mode, habitat use, and morphology in this assemblage of rap- tors; and (2) integrate these findings with re- spect to the question of morphological con- straints on behavior and habitat use. We also discuss the consequences of our findings for the future of these species in the face of continued habitat loss. METHODS Species and study areas.--Members of three of the five falconiform families (del Hoyo et al. 1994) occur in the Philippines (Dickinson et al. 1991). We were able to observe 21 species of 14 genera (Appendix), including three endemic species and nine endemic subspecies. Habitat analyses and behavioral observations were made during a field study over a period of 9.5 months (January to April 1993, November to Febru- ary 1994, and March to July 1994). Investigations were carried out in 19 study areas at elevations rang- ing from sea level to 2,500 m on the islands of Luzon (Sierra Madre, Quirino and Isabela; Mt. Makiling, Laguna), Mindanao (Mt. Kitanglad, Bukidnon; Car- men-Cantilan, Surigao del Sur; Mt. Apo, Davao City and North Cotabato; Initao, Misamis Oriental), and Palawan (El Nido and Bacuit Archipel, northern Pa- lawan). Observations in each particular area lasted from a few days to three weeks. Morphometry.--Study skins were measured using 384 specimens from various museum collections: British Museum (Tring), National Museum of Scot- land (Edinburgh), Rijks Museum van Natuurlijke Historie (Leiden, The Netherlands), Universitets Zoologiske Museum (Koberthavn, Denmark), Zool- ogisches Museum der Humboldt Universit;fit (Ber- lin), Staatliches Naturhistorisches Museum Braun- schweig (Brunswick, Germany), Naturhistorisches Museum Wien (Vienna), American Museum of Nat- ural History (New York), Smithsonian Institution (Washington, D.C.), Field Museum of Natural His- tory (Chicago), Museum of Natural History (Cincin- nati, USA), Delaware Museum of Natural History (Wilmington, USA), National Museum of the Phil- ippines (Manila), Wildlife Biology Laboratory at the University of the Philippines (Los Barios), and the Zoological Garden (Manila). Up to 39 study skins of adult specimens were an- alyzed for common species, whereas for rare species and subspecies, the number of individuals was much smaller. Sex ratios within samples of species varied between 0.45 and 0.55. External morphological fea- tures were measured mostly as defined in Leisler and Winkler (1985, 1991). The flight apparatus was rep- resented with 23 variables. These included length of all primaries; length of alula; lengths of the first, cen- tral, and innermost secondaries; lengths of outer and central rectrices; and wing length. Also used were the primary projection, or Kipp's distance, measured as the distance from the wing tip to the first second- ary; and tail gradation, the difference in length be- tween the innermost and outermost rectrices (Leisler and Winkler 1985, 1991). The mean correlation (ab- solute values) within this data set was 0.525 (range 0.007 to 0.991). Thirteen characters represented mor- phology of the hind limb. They comprised all toe lengths, talon chords, and diameters of talons at the base, and length of the tarsus. The mean correlation among these variables was 0.395 (range 0.005 to 0.977). The length, width, and depth of the bill were measured: (1) including the cere at the base of the bill, and (2) on the bill proper at the distal edge of the cere. The mean correlation among these six char- acters was 0.532 (range 0.006 to 0.935). Two additional characters, aspect ratio (wing span squared divided by wing area) and wing loading (body mass divided by wing area), were calculated from photographs of soaring birds. Data for the Phil- ippine Falconet (Microhierax erythrogenys) were taken from a freshly killed specimen. No suitable photo- graphs were available for the Philippine subspecies of Jerdon's Baza (Aviceda jerdoni magnirostris), so we used photographs of its closest equivalent, the Pacif- ic Baza (A. subcristata) from the Australian region (Marchant and Higgins 1993). We also had no pho- tographs of the Eastern Marsh Harrier (Circus spilon- otus). Photographs were scanned using the computer program Imagein 3.0, and contrast of the silhouettes was enhanced with program Coral Photopaint 3.0 to optimize image analysis with program Lucia 2.11. All measurements of soaring birds refer to fully out- stretched wings. Because the tail can be fanned to TABLE 1. Characteristics of the flight apparatus of 16 Philippine raptors and one ecological equivalent from Australia (in brackets). Linear- Wing Wing ized loading loading wing Body mass Body length w/o tail w/tail load- (g) (cm) Aspect ratio (N m 2) (N m 2) ing Species M F M F M F M F M F F [Aviceda subcristata] 307 347 Pernis ptilorhyncus 1,000 Pernis celebensis 679 738 Haliastur indus 450 581 Haliaeetus leucogaster 2,570 3,150 Spilornis cheela 688 853 Accipiter virgatus 95 154 Accipiter trivirgatus 199 353 Accipiter soloensis 156 204 Butastur indicus 375 433 Pithecophaga jefferyi 4,464 6,000 Hieraaetus kienerii 733 800 Spizaetus cirrhatus 1,360 1,600 Spizaetus philippensis -- 1,225 Microhierax erythrogenys 46 52 Falco tinnunculus 186 217 Falco peregrinus 650 1,000 40.0 5.64 5.95 20.6 25.0 17.6 20.1 0.209 58.0 65.0 5.68 5.61 34.4 30.8 30.0 23.4 0.179 53.5 55.4 5.32 5.41 26.0 26.0 22.4 22.4 0.175 47.1 48.9 6.26 6.45 17.2 21.3 15.6 19.6 0.171 70.0 77.0 7.67 7.25 44.4 46.3 43.0 42.6 0.209 52.4 54.5 5.26 5.29 25.5 28.9 22.5 25.6 0.185 25.7 28.1 4.72 4.34 21.1 28.4 16.5 22.2 0.247 33.2 37.5 5.75 5.77 25.4 35.9 19.7 27.7 0.241 27.8 29.0 5.68 6.14 27.3 33.1 22.1 26.9 0.256 42.6 42.9 6.34 6.44 27.2 30.7 22.0 25.8 0.216 95.2 105.2 3.89 3.83 48.7 51.0 42.2 44.3 0.178 46.8 52.6 5.61 5.68 33.2 31.6 27.1 26.1 0.200 58.4 64.9 5.48 5.35 44.9 41.9 37.2 33.5 0.214 56.6 58.2 4.98 4.55 -- 46.5 -- 38.5 0.203 15.9 17.2 5.74 5.66 25.4 26.4 21.9 22.9 0.217 33.8 36.2 7.07 6.92 24.1 24.5 18.0 18.4 0.268 48.9 52.8 5.61 6.13 47.5 57.2 39.1 46.8 0.266 various degrees, we used 60 ų as the standard tail-fan- ning angle for measurements of all species, and the silhouettes were modified accordingly. This value is the approximate average angle shown by soaring birds. Relevant values were calculated as described in Pennycuick (1975), Kerlinger (1989), and Norberg (1990). Wing loading (N m 2) was calculated accord- ing to Norberg (1990), either with the tail excluded or included. The two measures of wing loading differ widely in long-tailed species (Table 1). We also cal- culated linearized wing loading (Jaksi and Caroth- ers 1985), except that we used the sum of wing and body area instead of wing area alone. Body-mass data were taken from Bennet (1986), Goodman and Gonzales (1989), Dunning (1993), Marchant and Higgins (1993), and del Hoyo et al. (1994), and from labels attached to museum specimens. All measurements were taken as indices relative to body length (total body length minus tail length) be- cause we intended to analyze aspects of shape rather than size (Mosimann 1970), and all measurements were log-transformed. Even in raptors with pro- nounced sexual size dimorphism, morphological variation between the sexes is low compared with that among species (see Leisler and Winkler 1991). Therefore, we used species' means for all analyses. Field data.--Point observations (>2 h per observa- tion) were carried out at exposed sites, e.g. cliffs, clearings, and prominent trees. Through recurrent visits to each point at different times of the day, the observations covered the full daylight period (0600 to 1800). Observations made along line transects were carried out primarily for the study of birds within the forest interior. To maximize the number of different individuals, large areas (30 to 50 km 2) were investigated using both methods. These two methods have been used successfully in other stud- ies of tropical forest raptors (Thiollay 1989, Whiracre et al. 1992). Sample sizes ranged from 2 to 219 ob- servations per species. Only species with more than five observations were included in subsequent anal- yses. Behavioral and habitat variables were noted si- multaneously for each undisturbed hunting raptor at the first sighting. To assure as much statistical in- dependence as possible (see Hejl et al. 1990), the be- havior of raptors on the wing was recorded again only after five minutes had elapsed and, for perched individuals, additional data were taken only after the birds had changed their perches. These second- ary observations comprise only 1.1% of the total of 896 observations. Five behavioral traits associated with hunting were chosen: perching, powered flight (continuous flapping), flap-gliding, gliding, and soaring (Kerlin- ger 1989). Perching denotes a search mode during which the bird scans its surroundings from a perch. When gliding, the wings are held stretched out from the body in a fixed position without flapping, and the flight path is more or less straight. Soaring is a circular gliding mode used to gain altitude in ther- mals or slope updrafts. Powered flight comprises two forms, continuous flapping and flap-gliding (powerglide), a flight mode during which flapping is interspersed with gliding. The data were analyzed as arcsine square-root transformed percentages ex- pressing the proportion of these behavioral modes in the observations. To characterize habitat use by raptors, we estimat- ed forest canopy cover and the proportion (in 5% in- crements) of 16 habitat types (primary rain forest, secondary rain forest, degraded secondary forest, tree plantations, clearings, fallow land, grassland, pasture, banana, rice, other cultivation, road, settle- ment, water, coconuts, cliffs) within a 250-m radius of the location of every bird that we observed. Data were averaged for each species and then arcsine square-root transformed for analyses. Data analysis.--The data were evaluated using principal components analysis (PCA) and stepwise multivariate regression analysis. Principal compo- nents were based on correlation matrices. Particular- ly when many characters are used, there may be many relevant components judged by eigenvalues 1. In order to avoid over-interpretation of the data, however, we adopted the following criterion: a com- ponent was deemed to be relevant if its correspond- ing eigenvalue was greater than 1 in 95% of 10,000 bootstrap samples and explained more than 10% of the total variance. By applying this rule, we avoided discussing "significang' but biologically irrelevant components. PCA was used solely to describe rela- tionships within our data set, which was not intend- ed to represent a sample of a large, unknown statis- tical population. The results of the bootstrap analysis were also helpful in the interpretation of component loadings. Stepwise methods have been widely criti- cized (see James and McCulloch 1990), mainly be- cause they do not guarantee a selection of the most influential variables or even the best correlations. Consequently, we used stepwise multiple regression analysis mainly to show that a particular set of data can be used to predict the criterion variable; we have also tried to point out the biological relevance of the variables selected in each individual case. Although we cannot rule out the possibility that other variable combinations may produce better predictions, we are confident that our results provide good estimates for the strength of the association among ecological attributes, behavior, and morphology. Similarities due to phylogenetic inertia may cause statistical bias in correlational analyses (Felsenstein 1985, Harvey and Pagel 1991). Available phylogenet- ic analyses of raptors are widely divergent and do not include all of the taxa treated by us. Therefore, we constructed our own phylogenetic tree (Fig. 1) based on studies by De Boer and Sinoo (1984), Boyce and White (1987), Olsen et al. (1989), Marchant and Higgins (1993), Seibold et al. (1993), Helbig et al. (1994), Holdaway (1994), Griffiths (1994), Wink and Seibold (1995), and Wink et al. (in press). Independent contrasts were computed for the sta- tistical evaluation of stepwise multiple regression c Pe ptilorhyncus Pe celebensis I Avjerdoni Sp cheela F__- Ac virgatus I LAcsųlųensis Ac trivirgatus d Pijefferyi Sp cirrhatus Sp philippensls Hi kienerii Ha indus I Ha leucogaster Ci sp#onotus E Ci melanoleucos Bu indicus El caeruleus Fa tinnunculus -- Fa severus -- Fa peregrinus MI erythrogenys FIG. 1. Classification of Philippine raptors used to compute independent contrasts. analysis results. The variables were analyzed with multiple linear regression through the origin (Fel- senstein 1985). The independent contrasts were rath- er sensitive to the branch lengths. Therefore, in some cases we also present results with corrections that as- sume constant branch lengths between nodes. We used independent contrasts only as a means of re- ducing bias, and not as estimates for correlated evo- lutionary changes; that is, we did not necessarily as- sume that characters evolved homogeneously throughout the phylogeny of the group. All results of the regression analyses are presented as plots of the observed values of the criterion variable based on predictions from the multiple regression equations. To conserve the functional relationship, and to dis- play the positions of individual species in the scatter, we present the relations of the original variables and not those of the contrasts. All calculations were car- ried out with programs written by us. RESULTS Habitat.--Among the 21 raptors studied, 13 species occurred in habitats with more than 50% forest cover and therefore were defined as forest dwellers; the other 8 species occupied / PPPHk o Fti / eAte 1oo o 0 20 40 60 80 100 % forest cover FIC. 2. The relationship between height above ground at which a species forages and forest cover in Philippine raptors. Open circles denote species that hunt above the canopy, filled circles denote species that inhabit forest interior, and open triangles denote species that use open habitats. See Appendix for the abbreviations used to designate species. open habitats (Fig. 2). All forest dwellers except the Chinese Goshawk (Accipiter soloensis) are residents. Four species of the open and semi- open habitats are winter visitors only. The for- est dwellers fall into two distinct groups (Fig. 2): (1) four species that hunt inside the forest below the canopy (below 40 to 50 m), and (2) nine species that hunt within and above the canopy zone. Three principal components accounted for 68% of the total variation in habitat (n = 18 spe- cies). The first component (PC1) explained 34% of the variation (eigenvalue = 5.83) and de- scribed the contrast between densely forested habitats and open and extremely "disturbed" habitats consisting of grasslands and rice fields. PC2 explained 21% of the variation (ei- genvalue = 3.60) and was related to an increas- ing proportion of water (effectively, the sea), coconut palms, and rocky areas in the habitat versus degraded and secondary forest stands and banana cultivations. PC3 explained 13% of the variation (eigenvalue = 2.18) and described variation within cultivated areas, from settle- ments and neighboring farming areas to aban- doned fields. Much of the variation in habitat use concerned the open country birds, as de- fined above. Forest species formed a tight clus- ter, and those of the interior scored lower on the first principal component (t = 3.06, P = 0.011) but were not separated from those hunting mainly above the canopy in the second com- ponent (t = 1.68, P = 0.121). Behavior.--Two principal components ex- plained 81% of the overall variation in hunting mode (n = 18 species). PC1 explained 48% of the variation (eigenvalue = 2.57) and was as- sociated with the transition from sit-and-wait to flap-gliding hunting modes. PC2 explained 33% of the variation (eigenvalue = 1.26) and represented the contrast between soaring and the other flight modes. This suggests that in these birds, the difference between the use of different flight modes was nearly as important as the transition from active-search to sit-and- wait hunting. Morphology.--Three principal components accounted for 70.5% of the variation associated with flight and feeding apparatus. PC1 ex- plained 36% of the variation (eigenvalue = 15.25) and arranged the 21 species along a con- tinuum from species with narrow wings and less-pronounced notches to those with broad, deeply notched wings. The width of the wings was reflected in a positive correlation with length of the inner primaries (P1 to P6), the sec- ondaries, and the scapulars. PC2 combined characters of the flight apparatus with claw measurements and explained 20% of the total variance (eigenvalue = 8.51). With respect to the wing, this axis denotes a decrease in wing length that was mainly due to a reduced distal section of the wing and thus reduced point- edness of the wing (Kipp's distance), and to an increase in the number of wing slots. The claw measurements referred to increasingly large (length, diameter) inner and hind claws. Spe- cies scoring high on this axis are well equipped for killing comparatively large prey with their feet. PC3 explained 14% of the total variance (eigenvalue = 6.01) and was related to the feet and the bill; this component represented the major variation in the feeding apparatus to- gether with the "killing-grip variables." PC3 denoted an overall increase in the length of the toes together with lengthened outer and mid- dle claws. Bill size increased along this axis with a relatively smaller increase of the distal section and an almost parallel increase of bill width and height. This axis thus represents an increase in the relative size of the capture ap- paratus (feet) and bill that together are well suited for killing avian or insect prey. The sixth primary (P6) can be considered as Sp. Pj c.) P ('O"'ec/ ' _. Ato Sch OPpp -3 B Fti -6 -3 0 3 6 Prediction from morphology F]G. 3. Morphological correlates of habitat use in Philippine raptors. Results are from a multiple re- gression of morphological variables as predictors of habitat data based on PC1 scores. Values predicted from the regression equation are plotted on the X- axis, and observed scores are plotted on the Y-axis. Of the five morphological variables extracted, Kipp's distance and length of central rectrices negatively correlated with predicted habitat use, and diameter of hind claw, length of inner toe, and length of in- nermost seconda positively correlated with pre- dicted habitat use (see text). a key structure for the form of the wing. If P6 is long, the outer primaries (P7 to P10) are short, giving the wing a rounded appearance. If P6 is short, the outer primaries become lon- ger, and the wing is more slender and pointed. The length of P6 correlates well with PC1 (r = 0.89), and the lengths of P7 to P10 correlate negatively with PC2 (r = -0.47 to -0.79). Integration.--To identify morphological cor- relates of habitat use, we conducted a stepwise multiple regression analysis using PC1 of the habitat data against 42 morphological vari- ables. The five morphological variables extract- ed explained 87% of the total variance (R = 0.934; Fig. 3). Two characters (Kipp's distance, length of central rectrices) were negatively cor- related with predicted habitat use, whereas length of the inner toe, diameter of the hind claw, and length of the innermost secondary were positively correlated with predicted hab- itat use. The multiple R for the corresponding independent contrasts was 0.811 (P < 0.05). Thus, species that occurred in habitats with a higher percentage of forest cover tend to pos- sess square-cut tails and round, broad wings. They also possess hind limbs that are typical of raptors that prey on vertebrates. Stepwise multiple regression analysis of habitat using PC2 (forest interior vs. coastal) yielded a high multiple correlation (R = 0.900, P < 0.0001) with three morphological predic- tors. Wing length and tail gradation were pos- itively correlated, and length of outer rectrices was negatively correlated, with the tail becom- ing more wedge-shaped with increasing use of coastal habitats. When the clearly outlying White-bellied Sea-Eagle (Haliaeetus leucogaster) was excluded from analysis, the multiple cor- relation between the same variables and PC2 was still substantial (R = 0.785, P < 0.005), and the regression coefficients were similar. The difference between the observed PC2 score of the sea-eagle and that calculated from the re- gression equation is well within acceptable er- ror limits (deviation <90% of overall standard deviation of observations around the regres- sion). The R-value was 0.596 (P > 0.05) when computed from standardized independent con- trasts and 0.845 (P < 0.001) when computed with branch lengths set to equal. The most important structural habitat vari- able was canopy cover (cf. Thiollay 1996). Step- wise multiple regression analysis with canopy cover as the criterion variable produced a mul- tiple correlation of R = 0.906 (P < 0.0001), with Kipp's distance and length of central rectrices correlating negatively, and inner toe length and bill height correlating positively, with the pre- diction (Fig. 4). The corresponding R-value was 0.644 (P > 0.05; P < 0.05 using Kipp's distance and tail length alone) for the standardized in- dependent contrasts and 0.901 (P < 0.0005) for the independent contrasts calculated with branches set to equal length. As in the preced- ing analysis, these results indicate that raptors living in forested areas tend to possess round- ed wings. Species of the forest interior differed from those hunting within or above the canopy by being smaller (t = 3.30, P = 0.007; log-values of female body mass). No significant size differ- ences were found between the forest species and the open country species. We expected that the most important struc- tural habitat variable, canopy cover, would show the strongest association with flight mode. Stepwise multiple regression analysis 0 /0 Av I Bi .. I 2 3 4 Prediction from morphology FIG. 4. Morphological correlates of use of canopy cover by Philippine raptors. Results are from a mul- tiple regression of morphological variables as pre- dictors of habitat canopy cover. Values predicted from the regression equation are plotted on the X- axis, and observed scores are plotted on the Y-axis. Of the four morphological variables extracted, Kipp's distance, and length of inner remiges were negative- ly correlated with predicted canopy cover, and inner toe length and bill depth (without cere) were posi- tively correlated with predicted canopy cover (see text). (backward variable selection) showed that the search mode used in foraging predicted habitat use with respect to canopy cover very well. Perching, gliding, and soaring accounted for 81% of the variance in canopy cover (R = 0.900, P < 0.0001; R = 0.765, P < 0.01 for the stan- dardized independent contrasts). All coeffi- cients were positive; perching, however, was negatively correlated with the regression score. Basically, the results mean that soaring and gliding occurred mainly in connection with closed habitats. The plot of predicted versus observed values (Fig. 5A) shows that the spe- cies observed in totally open habitats, Eurasian Kestrel (Falco tinnunculus) and Eastern Marsh Harrier, may have determined the results. Thus, we repeated the analysis excluding these species in order to corroborate the results. The same variables were selected in the stepwise multiple regression analysis; the signs and magnitudes of the coefficients, as well as the correlations with the regression scores, were similar to the foregoing analysis. In particular, o 'A 0 I 2 3 ' , ZM, A sų I 2 3 4 Prediction from behavior Behavioral correlates of use of canopy FIG. 5. cover by Philippine raptors. Results are from a mul- tiple regression of behavioral variables as predictors of habitat canopy cover. Values predicted from the regression equation are plotted on the X-axis, and observed scores are plotted on the Y-axis. Three be- havioral variables were extracted: perching, gliding, and soaring (see text). (A) Analysis based on all spe- cies. (B) Analysis based on all species except Eur- asian Kestrel and Eastern Marsh Harrier. we obtained R = 0.783 (P < 0.01; R = 0.760, P < 0.025 for the independent contrasts), and the predicted scores for the two omitted species were very close to the observations (Fig. 5B). Stepwise multiple regression analysis was applied to test the correlation between modes of locomotion and morphological characters, including aspect ratio and wing loading (Table 1). First, we analyzed the relationship between 56 8ch 0 0 ),..S z/Pc$ 40 Hii /xH( Ppp ) HyAjm te 8 24 40 56 Prediction from morphology FIG. 6. Morphological correlates of soaring be- havior in Philippine raptors. Results are from a mul- tiple regression of morphological variables as pre- dictors of proportion of time soaring. Values pre- dicted from the regression equation are plotted on the X-axis, and observed scores are plotted on the Y- axis. All three of the morphological variables extract- ed (bill width without cere, tail graduation, and tar- sus length) were negatively correlated with propor- tion of time soaring (see text). 24.5 e 17.0 19.5 22.0 24.5 27.0 Prediction from behavior (PC) FIc. 7. Behavioral correlates of wing loading in Philippine raptors. Results are from a multiple re- gression of behavioral variables as predictors of lin- earized wing loading. Values predicted from the re- gression equation are plotted on the X-axis, and ob- served scores are plotted on the Y-axis. The two vari- ables extracted were scores of PC1 and PC2 of the behavioral data. The Pacific Baza (Asu) was used as a surrogate for ]erdon's Baza in this analysis (see text). morphology and soaring. A strong correlation (R = 0.878, P < 0.0001; R = 0.849, P < 0.001 for the independent contrasts) was found with three characters (all of them entered with a negative sign): (1) width of bill (without cere), (2) tail gradation, and (3) length of tarsus (Fig. 6). Raptors that soar frequently have a short tarsus, narrow bill, and a square-cut tail. With respect to the flight apparatus alone, the num- ber of primary notches was positively correlat- ed with soaring (r = 0.599, P < 0.01; r = 0.570, P < 0.05 for the corresponding independent contrasts). Wing loading and aspect ratio to- gether explained only 21% (P > 0.05) of the variance in soaring frequency. We also examined the claim by Jaksi and Ca- rothers (1985) that linearized wing loading is related to the sit-and-wait versus active-flight continuum of hunting mode. We used ordinary multiple regression, with the first two principal components derived from the behavioral data as predictors (Fig. 7). The correlation (R = 0.801, P < 0.001; R = 0.536, P > 0.05 for stan- dardized contrasts; R = 0.744, P < 0.01 for con- trasts with equal branch lengths) indicated that wing loading is low in large species in which foraging behavior is dominated by soaring in combination with a moderate proportion of sit- and-wait hunting. DIscussioN Our analyses revealed consistent compo- nents of morphological covariation with for- aging habitat and behavior. We also showed that the relationships among foraging behav- ior, habitat, and morphology to a large extent are independent of phylogenetic relationships. The shape variables that were the most impor- tant components of morphological variation pertain to characters that usually are not mea- sured at all--wing shape and the feet. Features of the bill figured only marginally in this vari- ation. One of the most popular characters in ecomorphological studies, tarsus length, did not contribute substantially to the three major components of morphological variation, and even wing length was not very important as a shape variable. The correlation analyses yield- ed similar results, although tarsus length was relevant for predicting the frequency of soaring (see below). Our findings stress how important it is to work with sufficiently large morpholog- ical character sets (Leisler and Winkler 1985, 1997), and they should be considered when dis- cussing results obtained in studies based on character sets restricted in number or on the functional complexes to which they belong (e.g. Schoener 1984, Hertel 1995). Several studies have discussed the flight style of raptors with respect to wing loading and aspect ratio (Pennycuick 1975, Rayner 1988, Jaksi and Carothers 1985, Janes 1985, Kerlinger 1989, Norberg 1981, 1990). These characters were not the best predictors of the frequency of certain flight modes in our study. Soaring, for instance, was best predicted by the number of notched primaries, and its close re- lationship with features of the feeding appa- ratus hints at the complex web of factors that may influence the evolution of behavioral and morphological traits. Wing loading and aspect ratio were not correlated significantly with soaring frequency. Although Rayner (1988) found it surprising that soaring hawks have low aspect ratios, subsequent discussions of the subject stressed that long wings may be dis- advantageous for taking off from perches and that thermal soaring requires a small turning radius (see Norberg 1990). Our findings con- firm the notion that for raptors that are limited in wing span (and hence have low aspect ratio and rounded wings), notched primaries are the appropriate alternative to reduce induced drag (Pennycuick 1972, Withers 1981, Kerlinger 1989, Norberg 1990). This, and the fact that low wing loading also is related to other aspects of flight performance (i.e. slow flight; Norberg 1990), may be responsible for the poor predic- tions derived from aspect ratio and wing load- ing alone. The hypothesis that raptors with high wing loading typically use sit-and-wait hunting modes, whereas species with low wing loading are active-search foragers (Jaksi and Carothers 1985), was supported by our data. However, light wing loading also was charac- teristic of most species for which sit-and-wait hunting was an important component of for- aging. Our correlative results concur with ideas de- veloped by Rochon-Duvigneaud (1952) and Voous (1969) on the functional and ecological importance of talon length, and they also stress that associated features, such as talon diameter and toe length, are relevant. Habitat, diet, and foraging mode may be considered independent niche dimensions (Price 1991), or they may be seen as completely congruent with each other. Although Rosenzweig (1985) adheres to the lat- ter view, Price (1991) showed that within Phyl- loscopus warblers, foraging behavior and habi- tat use largely are independent. Janes (1985) found no strong or consistent relationships be- tween habitat structure and the foraging be- havior of hawks. In our study, hunting behav- ior correlated well with habitat use, and the morphological predictors for habitat use and locomotion obtained with stepwise multiple regression revealed the complex interactions among these functions. We conclude that hab- itat use and foraging are neither completely equivalent nor uncorrelated with one another This is due, at least in part, to morphological constraints posed by conflicting demands of hunting, migration, and habitat use (Winkler and Leisler 1985, 1992). Forest-dwelling raptors are small and they have short, rounded wings as a general adap- tation to flight in dense vegetation. This results in a low aspect ratio and moderately high wing loading, and therefore a rather energetically ex- pensive flight mode (Jaksi6 and Carothers 1985, Kerlinger 1989, Norberg 1990). High-speed powered flight can be attained only by flying at a speed that is a large multiple of minimum- power speed (Pennycuick et al. 1994). Thus, a sit-and-wait strategy combined with short pur- suits is the most economical hunting mode for raptors in the forest interior but is not possible in open areas that lack suitable perches, even if prey are abundant (Preston 1990, Widen 1994). The accipiters are typical representatives of this group; only the migratory Chinese Gos- hawk, with the longest wing tip of the genus (Wattel 1973), hunts most frequently above the forest. Larger species are not able to hunt within densely vegetated areas. Therefore, they rely on the soaring mode of hunting, and they often perch in tree tops from which they launch their attacks. Their flight apparatus can be consid- ered a compromise among various constraints. The wings are not too long, because these rap- tors need to dive into the canopy (see also Brown 1976). Among the species that inhabitat open and semiopen habitats, the four winter visitors are completely independent of forest habitats. Species of open habitats, particularly migrants, are characterized by high aspect ra- tios and long, pointed wings. These morpho- logical specializations support, among other things, efficient long-distance flight (Winkler and Leisler 1992, Yong and Moore 1994, M6nk- k6nen 1995) and prevent these species from overlapping with species of the forest interior in morphology, behavior, and habitat require- ments (Thiollay 1985). In the Philippines, 13 species occur primarily in tropical rain forest, including the most en- dangered species observed during our field work, the Philippine Eagle (Pithecophaga jeffer- yi; Kennedy 1977). Eight of these species are ob- ligate inhabitants of mature rain forest. Their main prey, relatively large mammals and birds, also are restricted to these forest habitats. Some of these species also use secondary rain forests to some extent. These areas, however, are rath- er small and short-lived, and because of exten- sive exploitation, these forest remnants usually are converted within a few years into grass- lands. Owing to the deterioration of the envi- ronment through human activities (e.g. Salva- dor 1994, Thiollay 1994, 1996), all forest species are nearing extinction. Those residents needing only a minimum of forest have similar mor- phological characteristics developed as adap- tations for hunting over open surfaces, e.g. lakes, shores, or the air space high above the forest canopy. After deforestation, these spe- cies were able, to some extent, to exploit new habitat types such as grasslands or rice fields. Not actually threatened are the winter visitors. Members of these species find sufficient habitat in which they appear to thrive. That species of primary forests are the most threatened has been stated for Sumatra, involving in part the same or closely related species, and for other nearby tropical countries (Thiollay 1996). Our study sheds some light on the ecological rela- tionships underlying the conservation of these raptor species and how they were shaped by ecomorphological constraints. ACKNOWLEDGMENTS The study was supported by the Austrian Science Foundation (Fonds zur F6rderung der wissenschaft- lichen Forschung, Project P8889-BIO) and by the In- stitute for Wildlife Biology and Game Management (Agricultural University, Vienna). We are indebted to the Department for Environment and Natural Re- sources of the Republic of the Philippines, the Phil- ippine Eagle Conservation Program Foundation, the Haribon Foundation, Green Mindanao, N. Ingle (Da- vao), and the Technical Aid Agency of the Federal Republic of Germany for their cooperation on the Philippines. We also express thanks to the industrial concerns in Carmen (Puyat Logging) and Bislig (PRI), as well as to the local staff and our Philippine guides. The morphological portion of the study was made possible through the excellent cooperation of the museums and curators of the collections listed above. We are greatly indebted to A. Schuster, S. Teb- bich, and M. Zeiler for their support in the field, to J. Kickert for her advice on the computer-based image analysis, and to W. Pinsker for his support during many stages of the study. We also thank W. Bock, R. Dudley, and an anonymous reviewer for critical com- ments on the manuscript. LITERATURE CITED BENNET, P.M. 1986. Comparative studies of mor- phology, life history and ecology among birds. Ph.D. dissertation, University of Sussex, Brigh- ton, United Kingdom. BIERREGAARD, g. O. 1978. Morphological analysis of community structure in birds of prey. Ph.D. the- sis, University of Pennsylvania, Philadelphia. BOYCE, D. A., AND C. g. WHITE. 1987. Evolutionary aspects of kestrel systematics: A scenario. Pages 1-21 in The ancestral kestrel (D. M. Bird and R. Bowman, Eds.). Raptor Research Report No. 6. BROWN, L. H. 1976. Birds of prey, their biology and ecology. Hamlyn Publishing Group, Feltham, United Kingdom. BROWN L. H., AND D. AMADON. 1968. Eagles, hawks and falcons of the world. Country Life Books, London. CADE, t. J., AND R. D. DIGBY. 1982. The falcons of the world. Collins, London. COLLAR, N.J., M. J. CROSBY, AND A. J. STATTERSFIELD. 1994. Birds to watch 2: The world list of threat- ened birds. BirdLife International Conservation Series No. 4, Cambridge, United Kingdom. COLLINS, N.M., J. A. SAYER, AND T. C. WHITMORE. 1991. The conservation atlas of tropical forests: Asia and the Pacific. Macmillan, London. DE BOER, L. E. M., AND R. P. SINOO. 1984. A karyo- logical study of Accipitridae (Aves: Falconifor- mes), with karyotypic descriptions of 16 species new to cytology. Genetica (Dordrecht) 65:89- 107. DEL HOYO, J., A. ELLIOT% AND J. SARGATAL (Eds.). 1994. Handbook of the birds of the world, vol. 2. New World vultures to guineafowl. Lynx Edi- cions, Barcelona, Spain. DICKINSON, E. C., R. S. KENNEDY, AND K. C. PARKES. 1991. The birds of the Philippines. An annotated checklist. British Ornithologists' Union Check- list No. 12, London. DUNNING, J. B. 1993. CRC Handbook of avian body masses. CRC Press, Boca Raton, Florida. FELSENSTEIN, J. 1985. Phylogenies and the compara- tive method. American Naturalist 125:1-15. GOODMAN, S. g., AND P. C. GONZALES. 1989. Notes on Philippine birds. 12. Seven species new to Ca- tanduanes Island. Bulletin of the British Orni- thologists' Club 109:48-50. GRIFFITHS, C. S. 1994. Syringeal morphology and the phylogeny of the Falconidae. Condor 96:127- 140. HARVEY, P. H., AND g.D. PAGEL. 1991. The compar- ative method in evolutionary biology. Oxford University Press, Oxford. HELL, S. J., J. VERNER, AND G. W. BELL. 1990. Sequen- tial versus initial observations in studies of avian foraging. Studies in Avian Biology 13:166-173. HELBIG, A. J., I. SEIBOLD, W. BEDNAREK, P. GAUCHER, D. RISTOw, W. SCHARLAU, D. SCHMIDL, AND M. WN<. 1994. Phylogenetic relationships among falcon species (genus Falco) according to DNA sequence variation of the cytochrome b gene. Pages 593-599 in Raptor conservation today (B.- U. Meyburg and R. D. Chancellor, Eds.). Pica Press, East Sussex, United Kingdom. HERTEL, F. 1995. Ecomorphological indicators of feeding behavior in recent and fossil raptors. Auk 112:890-903. HOLDAWAY, g. N. 1994. An exploratory phylogenetic analysis of the genera of the Accipitridae, with notes on the biogeography of the family. Pages 601-637 in Raptor conservation today (B.-U. Meyburg and R. D. Chancellor, Eds.). Pica Press, East Sussex, United Kingdom. INTERNATIONAL COUNCIL FOR BIRD PRESERVATION. 1992. Putting biodiversity on the map: Priority areas for global conservation. ICBP, Cambridge, United Kingdom. JAKSI, F. 1985. Toward raptor community ecology: Behavior bases of assemblage structure. Journal of Raptor Research 19:107-112. JAKSI0_, E, AND J. H. CAROTHERS. 1985. Ecological, morphological, and bioenergetic correlates of hunting mode in hawks and owls. Ornis Scan- dinavica 16:165-172. JAMES, F. C., AND C. E. MCCULLOCH. 1990. Multivar- iate analysis in ecology and systematics: Pana- cea or Pandora's box? Annual Review of Ecology and Systematics 21:129-166. JANES, S. W. 1985. Habitat selection in raptorial birds. Pages 159-188 in Habitat selection in birds (M. L. Cody, Ed.). Academic Press, Orlando, Florida. JENKINS, g. R. 1995. Morphometrics and flight per- formance of southern African Peregrine and Lanner falcons. Journal of Avian Biology 26:49- 58. KARR, J. R., AND E C. JAMES. 1975. Eco-morphological configurations and convergent evolution in spe- cies and communities. Pages 258-291 in Ecology and evolution of communities (M. L. Cody and J. M. Diamond, Eds.). Belknap Press, Cam- bridge, Massachusetts. KEMP, A. C., ANO T. M. CROWE. 1994. Morphometrics of falconets and hunting behaviour of the Black- thighed Falconet Microhierax fringillarius. Ibis 136:44-49. KENNEDY, g. S. 1977. Notes on the biology and pop- ulation status of the Monkey-eating Eagle of the Philippines. Wilson Bulletin 89:1-20. KERLINGER, [9. 1989. Flight strategies of migrating hawks. University of Chicago Press, Chicago. LANDMANN, A., AND N. WINDING. 1993. Niche seg- regation in high-altitude Himalayan chats (Aves, Turdidae): Does morphology match ecology? Oecologia 95:506-519. LEDERER, R. J. 1984. A view of avian ecomorpholog- ical hypotheses. kologie der V6gel 6:119-126. LEISLER, B. 1980. Morphological aspects of ecological specialisation in bird genera. )kologie der V6- gel 2:199-220. LEISLER, B., H.-W. LEY, AND H. WINKLER. 1989. Hab- itat, behaviour and morphology of Acrocephalus warblers: An integrated analysis. Ornis Scandi- navica 20:181-186. LEISLER, B., AND H. WINKLER 1985. Ecomorphology. Current Ornithology 2:155-186. LEISLER, B., AND H. W1NKLER. 1991. Ergebnisse und Konzepte 6komorphologischer Untersuchungen an V6geln. Journal ftir Ornithologie 132:373- 425. LEISLER, B., AND H. W1NKLER. 1997. )komorpholo- gische Untersuchungen am Beispiel der Weber- v6gel (Ploceidae) und Eisv6gel (Alcedinidae). Annales Ornithologiae 21:17-43. MARCHANT, S., AND P. J. HIGGINS (Eds.). 1993. Hand- book of Australian, New Zealand and Antarctic birds, vol. 2. Raptors to lapwings. Oxford Uni- versity Press, Melbourne. MILES, D. B., AND R. E. RICKLEES. 1984. The correla- tion between ecology and morphology in decid- uous forest passerine birds. Ecology 65:1629- 1640. MILES, D. B., R. E. RICKLEES, AND J. TRAVIS. 1987. Concordance of ecomorphological relationships in three assemblages of passerine birds. Ameri- can Naturalist 129:347-364. MNKKSNEN, g. 1995. Do migrant birds have more pointed wings? A comparative study. Evolu- tionary Ecology 9:520-528. MOSIMANN, J. E. 1970. Size allometry: Size and shape variables with characterizations of the lognor- mal and generalized gamma distributions. Jour- nal of the American Statistical Society 65:930- 945. NIEBOER, E. 1973. Geographical and ecological dif- ferentiation in the genus Circus. Ph.D. thesis, University of Amsterdam, Amsterdam, The Netherlands. NIEMI, G. J. 1985. Patterns of morphological evolution in bird genera of New World and Old World peatlands. Ecology 66:1215-1228. NORBERG, U. M. 1981. Flight, morphology and the ecological niche in some birds and bats. Sym- posium of the Zoological Society London 48: 173-197. NORBERG, U. M. 1990. Vertebrate flight. Springer-Ver- lag, Berlin. OLSEN, P. D., R. C. MARSHALL, AND A. GAAL. 1989. Relationships within the genus Falco: A compar- ison of the electrophoretic patterns of feather proteins. Emu 89:193-203. PENNYCUICK, C. J. 1972. Soaring behaviour and per- formance of some East African birds, observed from a motor-glider. Ibis 114:178-218. PENNYCUICK, C. J. 1975. Mechanics of flight. Pages 1- 75 in Avian biology, vol. 5 (D. S. Farner and J. R. King, Eds.). Academic Press, New York. PENNYCUICK, C. J., M. R. FULLER, J. J. OAR, AND S. J. KIRKPATRICK. 1994. Falcon versus grouse: Flight adaptation of a predator and its prey. Journal of Avian Biology 25:39-49. PRESTON, C. R. 1990. Distribution of raptor foraging in relation to prey biomass and habitat structure. Condor 92:107-112. PRICE, T. 1991. Morphology and ecology of breeding warblers along an altitudinal gradient in Kash- mir, India. Journal of Animal Ecology 60:643- 664. RAYNER, J. M. V. 1988. Form and function in avian flight. Current Ornithology 5:1-66. ROCHON-DUvIGNEAUD, A. 1952. Les armes des ra- paces. L'Oiseau et la R.EO. 22:91-106. ROSENZWEIG, M. L. 1985. Some theoretical aspects of habitat selection. Pages 517-540 in Habitat selec- tion in birds (M. L. Cody, Ed.). Academic Press, Orlando, Florida. SALVADOR, D. J. I. 1994. Socio~economic incentives for the conservation of the rainforest habitat on the Philippine Eagle Pithecophaga jefferyi. Pages 277-282 in Raptor conservation today (B.-U. Meyburg and R. D. Chancellor, Eds.). Pica Press, East Sussex, United Kingdom. SCHOENER, t. 1984. Size differences among sympat- ric, bird-eating hawks: A worldwide survey. Pages 254-279 in Ecological communities: Con- ceptual issues and the evidence (D. R. Strong, D. S. Simberloff, and L. G. Abele, Eds.). Princeton University Press, Princeton, New Jersey. SEIBOLD, I., A. J. HELBIG, AND M. WINK. 1993. Molec- ular systematics of falcons (family Falconidae). Naturwissenschaften 80:87-90. STILES, F. G. 1995. Behavioral, ecological and mor- phological correlates of foraging for arthropods by the hummingbirds of a tropical wet forest. Condor 97:853-878. THIOLLAY, J.-M. 1985. Composition of falconiform communities along successional gradients from primary rainforest to secondary habitats. Pages 181-190 in Conservation studies on raptors (I. Newton and R. D. Chancellor, Eds.). ICBP Tech- nical Publication No. 5. Paston Press, Norwich, United Kingdom. THIOLLAY, J.-M. 1989. Censusing of diurnal raptors in a primary rainforest: Comparative methods and species detectability. Journal of Raptor Re- search 23:72-84. THIOLLAY, J.-M. 1994. A world review of tropical for- est raptors. Current trends, research objectives and conservation strategy. Pages 231-239 in Raptor conservation today (B.-U. Meyburg and R. D. Chancellor, Eds.). Pica Press, East Sussex, United Kingdom. THIOLLAY, J.-M. 1996. Rain forest raptor communi- ties in Sumatra: The conservation value of tra- ditional agroforests. Pages 245-261 in Raptors in human landscapes (D. M. Bird, D. E. Varland, and J. J. Negro, Eds.). Academic Press, London. Voous, K. H. 1969. Predation potential in birds of prey from Surinam. Ardea 57:117-148. WAINWRIGHT, P. C. 1991. Ecomorphology: Experi- mental functional anatomy for ecological prob- lems. American Zoologist 31:680-693. WATTEL, J. 1973. Geographical differentiation in the genus Accipiter. Publications of the Nuttall Or- nithological Club No. 13. WHITACRE, D. A., L. E. JONES, AND J. SUTTER. 1992. Censusing raptors and other birds in tropical forest: Further refinements of methodology. Pages 39-52 in Maya Project, Progress Report No. 5 (D. E Whitacre and R. K. Thorstrom, Eds.). The Peregrine Fund, Boise, Idaho. WIDI2N, P. 1994. Habitat quality for raptors: A field experiment. Journal of Avian Biology 25:219- 223. WIENS, J. A. 1989. The ecology of bird communities, vol. 1. Cambridge University Press, Cambridge, United Kingdom. WINK, M., AND I. SEIBOLD. 1995. Molecular phylog- eny of Mediterranean raptors (families Accipit- ridae and Falconidae). Pages 1-10 in Biologia y conservacion de las rapaces Mediterraneas (J. Muntaner and J. Mayol, Eds.). Monografias No. 4. SEO, Madrid. WINK, g., I. SEIBOLD, E LOTFIKHAH, AND W. BED- NAREK. Molecular systematics of Holarctic rap- tors. In press in Proceedings of the International Conference on Holarctic Birds of Prey. WINKLER, H. 1989. Verhalten als Mittler zwischen Morphologie und Habitat. Pages 345-350 in Cur- rent topics in avian biology. DO-G Meeting, Bonn, 1988 (R. van den Elzen, K.-L. Schuch- mann, and K. Schmidt-Koenig, Eds.). Deutsche Ornithologen-Gesellschaft, Garmisch-Parten- kirchen, Germany. WINKLER, H., AND g. LEISLER. 1985. Morphological aspects of habitat selection in birds. Pages 415- 434 in Habitat selection in birds (M. L. Cody, Ed.). Academic Press, Orlando, Florida. WINKLER, H., AND 1. LEISLER. 1992. On the ecomor- phology of migrants. Ibis 134 (Supplement):21- 28. WITHERS, P. C. 1981. The aerodynamic performance of the wing in Red-shouldered Hawk Buteo li- nearis and a possible aeroelastic role of wing-tip slots. Ibis 123:239-247. YONG, W., AND E R. MOORE. 1994. Flight morpholo- gy, energetic condition, and the stopover biology of migrating thrushes. Auk 111:683-692. Associate Editor: K. P. Dial APPENDIX. Species treated in this paper and their diets. Our observations augmented by data from Brown and Amadon (1969), Wattel (1973), Kennedy (1977), Cade and Digby (1982), H. Miranda (pers. comm.), Marchant and Higgins (1993), and del Hoyo et al. (1994). Diet: (?) = possibly, (o) = occasionally, (+) = regularly, (+ + ) = frequently, (+ + + ) = very frequently. Code Species Diet Hunting above the forest Ppp Pernis ptilorhyncus philippensis Oriental Honey-Buzzard Pcs Pernis celebensis steerei Barred Honey-Buzzard Pernis celebensis winkleri Sch Spilornis cheda holospilus Crested Serpent Eagle As Accipiter soloensis Chinese Goshawk Pj Pithecophaga jefferyi Philippine Eagle Hkf Hieraaetus kieneriiformosus Rufous-bellied Eagle Scl Spizaetus cirrhatus limnaetus Changeable Hawk-Eagle Sp Spizaetus p. philippensis Philippine Hawk-Eagle Spizaetus p. pinskeri Fpe Falco peregrinus ernesti Peregrine Falcon Hunting in the forest interior Ajn Aviceda jerdoni magnirostris Jerdon's Baza Av Accipiter virgatus confusus Besra Accipiter virgatus quagga Ate Accipiter trivirgatus extimus Crested Goshawk Me Microhierax e. erythrogenys Philippine Falconer Microhierax e. meridionalis Hunting outside the forest Ech Elanus caeruleus hypoleucus Black-shouldered Kite Hii Haliastur indus intermedius Brahminy Kite H1 Haliaeetus leucogaster White-bellied Sea-Eagle Cs Circus spilonotus Eastern Marsh Harrier Cm Circus melanoleucos Pied Harrier Bi Butastur indicus Grey-faced Buzzard Fti Falco tinnunculus interstinctus Eurasian Kestrel Fss Falco severus severus Oriental Hobby Wasp larvae (+ + + ), frogs (+), lizards (+), small nestlings (+) Wasp larvae (+ + +), frogs (?), lizards (+), small nestlings (+) Reptiles (+ + + ), small birds (+), small mammals (+ +) Insects (+), frogs (+ +), lizards (+), small birds (+), small mammals (+) Reptiles (+), medium and large birds (+ +), medium and large mammals (+++) Small and medium birds (+ +) Reptiles (+), medium and large birds (+ +), small to large mammals (+) Reptiles (o), small birds (+), me- dium birds (+ + + ), small mammals (+) Small birds (+ +), medium birds (+++) Insects (+ + + ), frogs (+), liz- ards (+) Small birds (+ + +), small mam- mals (+) Lizards (+), small birds (++), medium birds (+ +), small mammals (+) Insects (+ + + ), especially drag- onflies, butterflies and cicadas; small birds (o) Insects (+ + +), lizards (+), small mammals (+ + + ) Carrion (+ +), grasshoppers (+), fish (+), frogs (+), reptiles (+), small birds and mammals (+) Carion (o), fish (+ + +), reptiles (+), medium and large mama- mals (o) Carrion (o), insects (+), fish (o), frogs (++), lizards (+), small birds (+), small mammals (+ +) Insects (++), frogs (+), lizards (+), small birds (+), small mam- mals (+++) Fish (o), frogs (++), reptiles (+ + ), small mammals (+ + ) Insects (++), reptiles (+), small birds (o), small mammals (+ + +) Insects (+++), small birds (+ + +), bats (+ +)