Luca Luiselli, Francesco M. Angelici, and Godfrey C. Akani

Comparative feeding strategies and dietary plasticity of the sympatric cobras Naja melanoleuca and Naja nigricollis in three diverging Afrotropical habitats Luca Luiselli, Francesco M. Angelici, and Godfrey C. Akani 55 Abstract: Two cobra species are found in the forest block of southern Nigeria (West Africa). However, whereas the one species, the spitting cobra (Naja nigricollis), is often found in strongly altered habitats (including suburban areas), the other, the black forest cobra (Naja melanoleuca), is a typical forest species that is currently subject to a rapid decrease in population abundance because of intensive forest alteration and landscape modification in this part of Africa. We studied whether the body sizes, diets, and feeding strategies of these two species changed in relation to habitat type, and whether the ecological success of the one species versus the other in altered habitats depends upon greater dietary flexibility in prey type or prey size. Therefore, we divided our cobra records into three habitat categories: (1) suburbia, (2) plantation forest mosaic, and (3) mature forest. We observed that sexual size dimorphism was minor in both species and in all habitat types, and that intersexual differences in prey composition and prey size were also minor. Nevertheless, there was a remarkable ontogenetic change in taxonomic composition of the diet for one species (N. nigricollis, with juveniles taking almost exclusively lizards and adults taking small mammals, birds, and lizards) but not the other. Remarkably, the species that is less adapted to life in suburban areas showed a reduction in mean body size from the forest to suburbia, which may also indicate suboptimal adaptation to strongly altered habitats. Prey size was similar for the two species and in the three habitat types, and the relationships between prey size and predator size were similar. Thus, it seems unlikely that flexibility in prey-size patterns explains the greater colonizing success of N. nigricollis. Nevertheless, and although both species exhibited remarkable dietary flexibility, leading them to prey upon homeotherms as well as heterotherms and upon terrestrial as well as arboreal and even aquatic prey, there were important interspecific differences in prey composition that may explain the ecological success of N. nigricollis. The success of N. nigricollis likely lies not in dietary flexibility but in the consistency with which its juveniles prey upon a single prey type (lizards, mainly Agama agama) that is so abundant in nearly every altered habitat in Nigeria and is a virtually unlimited food resource for young N. nigricollis. However, adults of this species also forage frequently upon commensal rodents and poultry, which may also help it to colonize man-made habitats. Résumé : Deux espèces de cobras habitent la zone forestière du sud du Nigéria (Afrique 63 de l Ouest). Alors que l une des espèces (Naja nigricollis) se retrouve souvent dans les habitats modifiés (y compris les banlieues), l autre (Naja melanoleuca) est une espèce typique des forêts dont les populations subissent un déclin rapide à cause des changements importants de l habitat forestier et des modifications du paysage dans cette partie de l Afrique. Nous avons tenté de déterminer si la taille du corps, le régime alimentaire et les stratégies de recherche de nourriture des deux espèces ont changé en fonction du type d habitat et si le succès écologique de l une des deux espèces dans les habitats transformés peut être attribuable à une flexibilité alimentaire plus grande quant au type de proies et à leur taille. Nous avons réparti nos données selon trois catégories d habitat : (1) les banlieues, (2) la mosaïque plantations forêts et (3) les forêts en pleine maturité. Le dimorphisme sexuel de la taille est faible chez les deux espèces et dans tous les types d habitat; les différences entre mâles et femelles quant au choix et à la taille de leurs proies sont peu importantes. Malgré cela, nous avons observé un changement ontogénique remarquable de la composition taxonomique du régime alimentaire chez l une des espèces (les juvéniles de N. nigricollis consomment presque exclusivement des lézards, alors que les adultes consomment des petits mammifères, des oiseaux et des lézards), mais pas chez l autre. Curieusement, chez l espèce la moins bien adaptée aux banlieues, la taille moyenne du corps diminue des forêts aux banlieues, ce qui peut représenter une adaptation sub-optimale aux habitats fortement modifiés. La taille des proies, ainsi que la relation entre la taille des proies et celle des prédateurs, sont semblables chez les deux espèces et dans les trois types d habitat. Received 12 April 2001. Accepted 13 September 2001. Published on the NRC Research Press Web site at http://cjz.nrc.ca on 17 January 2002. L. Luiselli 1 and F.M. Angelici. F.I.Z.V. (Italian Foundation of Vertebrate Zoology), via Olona 7, I-00198 Rome, Italy, and Institute of Environmental Studies Demetra, via dei Cochi 48/B, I-00133 Rome, Italy. G.C. Akani. Department of Biological Sciences, Rivers State University of Science and Technology, P.M.B. 5080, Nkpolu, Port Harcourt, Rivers State, Nigeria. 1 Corresponding author (e-mail: lucamlu@tin.it). Can. J. Zool. 80: 55 63 (2002) DOI: 10.1139/Z01-178

56 Can. J. Zool. Vol. 80, 2002 Il semble donc peu probable que la flexibilité dans le choix des proies puisse expliquer le succès plus grand de N. nigricollis. Néanmoins, même si les deux espèces démontrent une flexibilité alimentaire remarquable, consommant des homéothermes aussi bien que des hétérothermes, des animaux terrestres, arboricoles et même parfois aquatiques, les différences inter-spécifiques importantes dans la composition des proies peuvent expliquer le succès écologique de N. nigricollis. Plutôt que la flexibilité alimentaire, c est probablement la continuité de l utilisation d un seul type de proie (des lézards, surtout Agama agama), qui explique le succès de N. nigricollis; cette espèce de lézard est abondante dans presque tous les habitats modifiés du Nigéria et elle constitue une ressource pratiquement illimitée de nourriture pour les jeunes cobras. Cependant, les adultes consomment également des rongeurs commensaux et de la volaille, ce qui peut les aider à coloniser les habitats anthropiques. [Traduit par la Rédaction] Introduction Luiselli et al. Nowadays, the overwhelming pressure of humans on the natural environment is leading to the destruction of large portions of pristine habitat and the formation of new types of habitats, so animals must readapt in order to survive. It is thus especially interesting to study the dynamics of adaptation of species to these newly available habitats. The ability to efficiently colonize new environments is clearly a crucial component of the ecological success of a given species, and is therefore a central issue in the study of conservation, evolution, and adaptation. In the ecological literature there is much empirical evidence that dietary plasticity may considerably help a given predator to colonize habitats different from those originally inhabited by its conspecifics, but nearly all case studies of vertebrates have been of birds or mammals (e.g., see Lovari et al. 1976; Moulton and Pimm 1983; Ehlrich 1989; Pimm 1991; Brandon 1995; Brooke et al. 1995; Case 1996; Williamson 1996; Altmann 1998; Capizzi 2000), whereas reptiles, particularly snakes, have been less studied in this respect (but see Petren and Case 1998; Losos and Spiller 1999). In this paper we use two species of large, highly venomous snakes for a case study: the spitting cobra (Naja nigricollis) and the black forest cobra (Naja melanoleuca). In particular, we analyse the dietary habits of these species in natural and man-made habitats, and try to determine whether the ecological success of the one species versus the other in man-made habitats depends on greater plasticity (in both prey type and prey size) in foraging mode. We conducted our study in a region of tropical Africa (southern Nigeria) where the formation of human-derived habitats is a particularly relevant ecological trend. In fact, nowadays the former rain forest (belonging to the Guinea Congo belt) is extremely fragmented, and is largely replaced with urban centres, plantations, and farmlands (De Montclos 1994); even an enormous portion of Nigeria s formerly forested territory is now a new man-made major vegetation zone, the so-called derived savanna (White 1983). Why should we care about using cobras as a case study to illustrate the above-mentioned issue? Apart from the practical implications of the study (for example, cobra bites are often lethal for humans, and a better knowledge of their ecology would allow the pertinent authorities to design proper plans for reducing the negative impact of these snakes on people in man-made habitats), cobras in southern Nigeria offer an ideal model for studying our issue. In fact, while the two cobra species are apparently similar in several ecological attributes (including a certain dietary generalism; Luiselli and Angelici 2000), they clearly differ in terms of their habitat requirements: the forest cobra is mainly linked to mature forests and swamp-forests and rarely enters very altered habitats, whereas the spitting cobra is often found in highly disturbed habitats, including suburbia, plantations, farmlands, and derived savannas (Luiselli and Angelici 2000). Thus, although there are sites where the two species are syntopic, it seems that N. melanoleuca has benefited from extensive forest alteration, and has reached a wider distribution in Nigeria than in the past, possibly challenging N. melanoleuca in its niche (cf. Luiselli and Angelici 2000 and references therein). It is also apparent that N. melanoleuca, once the most common venomous snake of the region, is now suffering from both forest habitat loss and intrusion of N. melanoleuca into areas where it was not present before, and is becoming rare or extirpated in wide areas where it was very common just a few decades ago (Politano 1997; Luiselli and Angelici 2000). Given that the two species are similar in terms of general ecological attributes, it is relevant to ask why N. melanoleuca exploits recently deforested areas better than N. melanoleuca does. Materials and methods The field study was carried out from September 1996 to September 2000 (though some data were recorded in 1994 and 1995) in several localities in southeastern Nigeria (for the territories surveyed see Luiselli and Angelici 2000; for a detailed description of the environment see Luiselli et al. 2000a). The study region is tropical, with the wet season from May to September and the dry season from October to April. The wettest period of the year is June July and the driest period is between late December and February. Activity of the two cobra species peaks during the wet season (Luiselli and Angelici 2000). The methods used to survey the study area, capture cobras, and analyse their food items are fully detailed in Luiselli and Angelici (2000); here we present only a brief summary. Fieldwork was conducted under all climatic conditions, but with a bias towards daylight hours (from 8 a.m. to 6 p.m.), owing to security constraints related to the prevailing unstable political situation. Field effort was almost identical in the wet and dry seasons: 378 and 391 days, respectively. Random routes throughout every available macrohabitat type in each study area were followed to locate snakes (for a characterization of the various study habitats see Luiselli and Angelici 2000). When seen, the cobras were captured by

Luiselli et al. 57 Table 1. Total lengths (cm; mean ± 1 SD) of cobras captured in three habitat types in southern Nigeria. Species and habitat type Males Females Naja nigricollis Suburbia 126.3 ± 18.3 (N = 31) 121.7 ± 13.4 (N = 38) Plantation forest mosaic 124.7 ± 11.8 (N = 34) 123.6 ± 10.6 (N = 23) Mature forest 123.3 ± 14.3 (N = 5) 125.0 ± 6.7 (N =5) Naja melanoleuca Suburbia 124.9 ± 6.4 (N = 4) 127.1 ± 8.3 (N =6) Plantation forest mosaic 154.8 ± 14.8 (N = 34) 153.3 ± 11.6 (N = 19) Mature forest 163.8 ± 19.8 (N = 35) 165.3 ± 19.1 (N = 25) Note: Only specimens >100 cm total length are included in the table. For statistical details see the text. hand with the aid of sticks, but additional free-ranging specimens were captured in pitfall traps with drift fences and in traps used by local people to capture frogs and fish. The site of capture and the habitat at each capture site were recorded. The captured snakes were sexed and measured for snout vent length (SVL; to the nearest 1 cm) and tail length. Then the abdomen was palpated until regurgitation of ingested food or defecation occurred. Specimens found already dead (road-killed or macheted by farmers or preserved in local collections; for a list of these see Luiselli and Angelici 2000) were dissected to obtain diet data. No specimens were killed or damaged for the purposes of this study. Prey items were identified to the lowest taxon possible. The mass of the prey item at the time of its ingestion was estimated, when possible, by comparing the item with intact conspecifics of various sizes from the authors personal collection, or determining the fresh biomass of perfectly preserved items. This was not possible for fecal items, which generally consisted of scales of reptiles and fur of mammals. For analysing cobra diets in the various habitats, using maps (scale 1:10000) of the vegetation and landscape of the surveyed territory produced by the Federal Government of Nigeria and (or) Aquater S.p.A. for environmental assessment reports, we assigned each snake record to one of three groups according to their habitat of capture: (1) specimens captured in suburbia and wide plantations or derived savannas, i.e., areas with <15% of forested territory within a 10-km radius of the site of capture; (2) specimens captured in plantation forest mosaics, i.e., areas with <65 and >15% of forested territory within a 10-km radius of the site of capture; and (3) specimens captured in mature forests, i.e., areas with >65% of forested territory within a 10-km radius of the site of capture. It should be mentioned that forested patches in groups 1 and 2 were secondary or at least partially altered forests, whereas those in group 3 were often undisturbed or slightly altered forests. Specimens captured in mangroves were also classified in one of the three groups depending on the status and surface of the mangrove formations, i.e., mangrove formations were considered in exactly the same way as forest types. To avoid a strong bias in data collection, as far as possible we maintained similarity of field effort in each of the three habitat groups. Specimens <100 cm in total length were considered juveniles (cf. Luiselli and Angelici 2000). Statistical analyses were performed with all tests twotailed and α set at 5%. Data on raw body masses of both Fig. 1. Percentages of specimens of two species of cobra, Naja melanoleuca (N = 142) and N. nigricollis (N = 184), captured in three habitat types in southeastern Nigeria, 1994 2000. snakes and their prey were natural-log-transformed to meet the assumptions of linearity. Yates correction was applied to contingency-table analyses when appropriate. In the text we present values as the mean ± 1 standard deviation. Although we included food items from recaptured specimens in the analyses, we exclude the possibility that pseudoreplication may have biased the statistical results because (i) the recapture entries were few and (ii) all of them occurred over a long time span (at least 3 months from one capture to the next). Results Sample sizes When the three study habitats were pooled, and both captures and recaptures included, 203 N. nigricollis and 197 N. melanoleuca were examined to obtain dietary data. Excluding recaptures, 96 N. nigricollis (31 males, 38 females, 27 juveniles) were captured in suburbia, 77 in plantation forest mosaics (34 males, 23 females, 20 juveniles), and 11 in mature forests (5 males, 5 females, 1 juvenile; Fig. 1). Means and standard errors of total lengths (not including those of juveniles) of these cobras in the three habitat types are presented in Table 1. Kruskal Wallis ANOVAs indicated

58 Can. J. Zool. Vol. 80, 2002 Fig. 2. Percentages of specimens of two cobra species, N. melanoleuca (N = 142) and N. nigricollis (N = 184), captured in three habitat types in southeastern Nigeria between 1994 and 2000 that contained prey. that size dimorphism was not significant either between the sexes or among habitat types (at least P > 0.1 in both cases). Excluding recaptures, 13 N. melanoleuca (4 males, 6 females, 3 juveniles) were captured in suburbia, 61 in plantation forest mosaics (34 males, 19 females, 8 juveniles), and 68 in mature forests (35 males, 25 females, 8 juveniles; Fig. 1). Mean total lengths (not including those of juveniles) in the three habitats are presented in Table 1. Kruskal Wallis ANOVAs indicated that size dimorphism was not significant between the sexes (P > 0.1), but was highly significant among habitat types (P < 0.0001), and a Tukey s HSD post-hoc test revealed that both males and females attained significantly larger sizes in plantation forest mosaics and mature forests than in suburbia (P < 0.0001), and even attained a significantly larger size in mature forests than in plantation forest mosaics (P < 0.001). To exclude the possibility that suburban N. melanoleuca were smaller than conspecifics from mature forests and plantation forest mosaics merely because the samples were of different ages (i.e., we captured, on average, younger specimens in suburbia than in the other two habitat types), we compared the body condition of the three cobra samples. To do this, we ln-transformed SVL and body mass of the various individuals and then did an analysis of covariance on the slopes of the regressions between these variables. It indicated significant differences in the slopes of the regression lines relative to the three habitats (at least P < 0.01 for all comparisons), the specimens from the suburban habitat having less mass for the same body length. It clearly indicated that suburban cobras were in poorer body condition than conspecifics from other habitats, i.e., they were not younger. Feeding frequency Percentages of snakes captured with prey in the different habitat types are shown in Fig. 2. Contingency-table analysis revealed that in every habitat type, the proportions of N. melanoleuca that had fed significantly exceeded the proportions of N. nigricollis that had fed (Yates corrected χ [] 1 2 test, at least P < 0.05 in all cases). Table 2. Prey eaten by cobras in suburbia in southern Nigeria. N. nigricollis N. melanoleuca Prey M F J M F J Mammals Rattus rattus 1 1 1 2 Mus musculoides 3 3 1 2 Birds Poultry (eggs and chicks) 3 6 1 1 Undetermined 1 1 Reptiles Agama agama 6 10 12 1 Mabuya affinis 2 1 Amphibians Ptychadena sp. Adults 3 2 2 1 Tadpoles 2* 2* 2* Bufo sp. (tadpoles) 1* 3* 1* Note: M, males; F, females; J, juveniles. *An undetermined number of tadpoles were eaten by 3, 3, and 1 N. melanoleuca (Ptychadena sp. tadpoles) and by 1, 3, and 1 N. melanoleuca (Bufo sp. tadpoles). In all cases the consumed material was too digested to determine the exact number of tadpoles eaten. Food habits Suburbia We removed 53 identifiable prey items from N. nigricollis (16 from males, 22 from females, 15 from juveniles) and 14 from N. melanoleuca (7 from males, 5 from females, 2 from juveniles) plus an undetermined number of tadpoles from several specimens of the latter species (Table 2). Endotherms and ectoterms were consumed by both cobra species, but important interspecific differences in taxonomic dietary composition emerged. When specimens >100 cm SVL are considered, N. nigricollis preyed most frequently on commensal rodents, poultry (both eggs and chicks), agamid lizards, and semiaquatic frogs. Adult N. melanoleuca also fed on rodents, birds, and amphibians, and preyed frequently upon anuran tadpoles but did not consume lizards. Moreover, whereas adult N. nigricollis showed a clear preference for given prey types (chiefly agamids, which accounted for 42.1% of the total number of items (N = 38) and poultry (which accounted for 23.7%)), adult N. melanoleuca had a more variable diet, showing a preference for frogs. If we look at juveniles, this interspecific divergence is even more evident (Table 2): nearly all N. nigricollis preyed upon agamids (85.7% of the total items, N = 14), whereas young N. melanoleuca were found to feed with similar frequency upon anuran tadpoles and lizards (both Scincidae and Agamidae). Plantation forest mosaic In total, 53 prey items were removed from N. nigricollis (20 from males, 16 from females, 17 from juveniles) and 62 from N. melanoleuca (22 from males, 28 from females, 12 from juveniles); also, an undetermined number of tadpoles were removed from 4 juveniles of the latter species (Table 3). In both species some specimens had multiple prey in the stomachs. Naja nigricollis >100 cm SVL fed mainly upon rodents (especially the striped rat, Lemniscomys striatus)

Luiselli et al. 59 Table 3. Prey eaten by cobras in plantation forest mosaics in southern Nigeria. N. nigricollis N. melanoleuca Prey item M F J M F J Mammals Mus musculoides 1 2 1 1 3 Lemniscomys striatus 6 7 4 9 2 Malacomys longipes 1 1 Funisciurus sp. 1 1 Undetermined rodents 4 3 Crocidura sp. 2 Birds Undetermined passerine 2 Reptiles Lamprophis fuliginosus 1 Psammophis sp. 1 Natriciteres sp. 1 Agama agama 3 1 8 Mabuya spp. 6 3 4 1 Mochlus fernandii 1 Panaspis sp. 1 Undetermined lizards 2 Amphibians Ptychadena sp. adults 2 3 1 2 2 1 Hoplobatrachus occipitalis 1 Undetermined Ranidae 1 3 Bufo sp. 1 Undetermined frog 4 2 1 Tadpoles (undetermined species) 4* Fishes Periophtalmus sp. 2 2 3 Note: M, males; F, females; J, juveniles. *An undetermined number of tadpoles were eaten by 0, 0, and 4 N. melanoleuca. In all cases the consumed material was too digested to determine the number of tadpoles eaten. and lizards (both Scincidae and Agamidae), whereas the diet of juveniles consisted mainly of agamids and skinks. In contrast, adult N. melanoleuca preyed mainly upon rodents and amphibians, and very occasionally upon birds and squamate reptiles. Thus, a clear-cut interspecific difference was found in terms of taxonomic dietary composition, although both species exhibited a wide dietary spectrum in terms of prey types. Mature forest A list of prey items found in the guts of both species in mature forest habitat is presented in Table 4. Because of the rarity of N. nigricollis in this habitat (see Luiselli and Angelici 2000), we were able to collect only 15 dietary records from this species (4 from males, 4 from females, 7 from juveniles). On the other hand, we collected 76 prey items from N. melanoleuca: 35 from males, 31 from females, and 10 from juveniles (Table 4). The bulk of the diet of adult N. nigricollis consisted of rodents, whereas adult N. melanoleuca fed with similar frequency upon rodents, frogs, and fish. With regard to juveniles, the few young N. nigricollis had fed upon lizards and fish (Periophthalmus spp.), whereas young N. melanoleuca had preyed upon a wide variety of appropriately sized organisms (20 100 g mass) from fishes to mammals. Table 4. Prey eaten by cobras in mature forests of southern Nigeria. N. nigricollis N. melanoleuca Prey item M F J M F J Mammals Dendromus sp. 1 Lemniscomys striatus 2 1 2 4 Lophuromys sp. 1 1 Undetermined Muridae 2 3 11 7 3 Birds Undetermined passerine 1 1 Reptiles Undetermined Scincidae 3 1 Amphibians Bufo sp. 1 1 Undetermined frogs 4 4 2 Fishes Periophthalmus sp. 2 7 7 3 Undetermined Cyprinidae 3 2 1 Undetermined fishes 2 4 4 Note: M, males; F, females; J, juveniles.

60 Can. J. Zool. Vol. 80, 2002 Fig. 3. Variations in percentages of taxonomic diet composition of the two cobra species, N. nigricollis (a) and N. melanoleuca (b), in relation to the three habitat types. The major prey categories (mammals (M), birds (B), reptiles (R), amphibians (A), and fish (F)) are used, and in the case of the tadpoles eaten (included in amphibians in this figure), the numbers of snakes containing that prey type were considered, given that it was impossible to determine the exact number of tadpoles ingested at every meal. Interhabitat comparisons The variations in the percent taxonomic composition of the diet of the two species of cobras in relation to the three habitat types are presented in Fig. 3. For both species the general composition of the diet changed significantly from one habitat to another (χ 2 test, at least P < 0.0001 for both species). When we looked at the data in more detail, it appeared that in N. nigricollis the important habitat-dependent dietary variations concerned reptiles (always the dominant prey type except in mature forests), fish (found only in mature forests), and mammals (the dominant prey type in mature forests, more rarely eaten in the other two habitats). With regard to N. melanoleuca, the most important habitatdependent dietary variations concerned amphibians (especially in plantation forest mosaics) and fish (important in mature forests), whereas mammals, birds, and reptiles accounted for fairly constant proportions of this species dietary spectrum. Nevertheless, it is noteworthy that samples of N. nigricollis from mature forests and N. melanoleuca from suburbia were quite small, which may have partially biased the data in these two cases. The intersexual differences in prey composition were never significant for the two species or for the three habitat types (χ 2 test with Yates correction, at least P > 0.08 in all cases). However, the composition of the juveniles diet was significantly different from that of the adults in N. nigricollis in suburbia and plantation forest mosaics (χ 2 test, at least P < 0.01 in both cases; lizards dominated the diets of immature snakes, but adults diets contained a wide range of prey types), whereas the sample from mature forests was too small to make any sound comparison. With regard to N. melanoleuca, we were unable to detect significant age-related differences via χ 2 test in both plantation forest mosaics and mature forests, whereas the sample studied in suburbia was very limited and impeded any reliable comparison. Prey size predator size relationships Suburbia Prey mass was measured with precision during 16 predation events concerning N. nigricollis (snake mass = 502.9 ± 303.9 g; range = 69 865 g) and 6 predation events concern-

Luiselli et al. 61 ing N. melanoleuca (snake mass = 509.5 ± 393.2 g; range = 71 1043 g). Prey mass did not differ significantly between cobra species (snake mass = 47.2 ± 47.1 g (range = 3 163 g) for N. nigricollis and 40.2 ± 63.6 g (range = 3.2 168.1 g) for N. melanoleuca) (t [20] = 0.29, P = 0.778); also, the mean body mass of the examined cobra samples did not differ significantly between the species (t [20] = 0.04, P = 0.967). Nevertheless, N. nigricollis tended to prey upon slightly larger prey in relation to their own body size than N. melanoleuca (mean prey mass/predator mass = 0.094 for N. nigricollis and 0.079 for N. melanoleuca), but the interspecific difference did not attain statistical significance in a two-tailed t test. In both cobra species there was a significant positive relationship between ln-transformed prey mass and ln-transformed predator mass (N. nigricollis: adjusted r 2 = 0.6501, slope = 1.002 ± 0.196, y-intercept = 2.583 ± 1.178; N. melanoleuca: adjusted r 2 = 0.8339, slope = 1.148 ± 0.256, y-intercept = 3.886 ± 1.512), and the slope deviated significantly from zero in both cases (F = 26.02, P = 0.0002, for N. nigricollis and F = 20.08, P = 0.011, for N. melanoleuca). The regression line of the prey mass predator mass relationship relative to N. nigricollis did not differ from that of N. melanoleuca (heterogeneity of slopes: F [1,18] = 0.2056, P = 0.656; heterogeneity of y-intercepts: F [1,19] = 1.9787, P = 0.176). Since the slopes and y-intercepts were not significantly different, it is possible to calculate the pooled slopes and the pooled y-intercepts. The pooled slope was 1.05573 and the pooled y-intercept was 3.02465. Plantation forest mosaic We were able to measure with precision the mass of the prey ingested by 21 N. nigricollis (snake mass = 650.4 ± 307.1 g (mean ± SD); range = 88 1142 g) and 23 N. melanoleuca (snake mass = 1228.5 ± 935.2 g; range = 98 3148 g). Prey mass of N. nigricollis (38.0 ± 37.5 g (mean ± SD); range = 5 162.8 g) was significantly less than that of N. melanoleuca (78.7 ± 79.2 g; range = 5.8 306.1 g) (t [42] = 2.15, P = 0.038), and also the mean mass of the examined cobra samples differed significantly between the species (t [42] = 2.71, P = 0.0099). Nevertheless, the mean prey mass/predator mass ratio was similar for the two cobra species (0.058 in N. nigricollis and 0.064 in N. melanoleuca) (two-tailed t test with df = 42, P = 0.875). In both cobra species there was a significant positive relationship between ln-transformed prey mass and ln-transformed predator mass (N. nigricollis: adjusted r 2 = 0.1886, slope = 0.5195 ± 0.247, y-intercept = 0.00594 ± 1.566; N. melanoleuca: adjusted r 2 = 0.6962, slope = 1.005 ± 0.1449, y-intercept = 2.994 ± 0.998), and slope deviated significantly from zero in both cases (F = 4.416, P = 0.0492, in N. nigricollis, and F = 48.12, P < 0.0001, in N. melanoleuca). The regression line for N. nigricollis did not differ from that for N. melanoleuca (heterogeneity of slopes: F [1,40] = 3.085, P = 0.086; heterogeneity of y-intercepts: F [1,41] = 0.329, P = 0.569). The pooled slope value was 0.840471 and the pooled y-intercept was 1.94493. Mature forest We recorded prey and predator sizes for 9 N. nigricollis (snake mass = 640.3 ± 272.8 g; range = 196 1033 g) and 26 N. melanoleuca (snake mass = 841.8 ± 700.4 g; range = 111 2514 g). Prey mass for N. nigricollis (36.6 ± 18.6 g) was similar to that for N. melanoleuca (34.8 ± 22.4 g) (t [33] = 0.83, P = 0.410), and the mean predator mass for the examined cobra samples was similar for the two species (t [33] = 0.22, P = 0.831). Nevertheless, the mean prey mass/predator mass ratio was significantly higher in N. nigricollis (0.057 versus 0.041) (two-tailed t test with df = 33, P = 0.04). In both cobra species there was a significant positive relationship between ln-transformed prey mass and ln-transformed predator mass (N. nigricollis: adjusted r 2 = 0.434, slope = 0.5913 ± 0.255, y-intercept = 0.2641 ± 1.625; N. melanoleuca: adjusted r 2 = 0.149, slope = 0.297 ± 0.148, y-intercept = 1.464 ± 0.949), but the slopes did not significantly deviate from zero for either species (F = 5.376, P = 0.0535, in N. nigricollis and F = 4.017, P = 0.057, in N. melanoleuca). The regression line for N. nigricollis did not differ from that for N. melanoleuca (heterogeneity of slopes: F [1,30] = 0.4919, P = 0.488; heterogeneity of y-intercepts: F [1,30] = 0.3789, P = 0.543). The pooled slope was 0.32768, and the pooled y-intercept was 1.30472. Interhabitat comparisons In general, both cobra species tended to take relatively small prey (in relation to their own body size) in every habitat type (mean prey mass/predator mass ratios ranged from 0.057 to 0.094 for N. nigricollis and from 0.041 to 0.079 for N. melanoleuca). Nevertheless, there was a clear trend, significant in both species (Kruskal Wallis ANOVA with Tukey s HSD post-hoc test, P < 0.01), to feed upon disproportionately larger prey in suburbia than in the other two habitat types, and upon relatively small prey in mature forests (the latter trend was very evident only in N. melanoleuca). The consequence of the above-mentioned patterns is that, whereas in the two species in both suburbia and plantation-forest mosaics the slope of the regression of prey mass against predator mass deviated significantly from zero, this was not the case for cobras of either species in mature forests. In fact, the pooled slope of the regression relating to mature forests was significantly less inclined than that of the regressions relating to the other two habitat types (heterogeneity of slopes: F = 21.6317, P = 0.0088). Nevertheless, in this case also, the two cobra species exhibited similar patterns of prey size predator size relationships, so the variation in the studied system depended more on the habitat type than on the species. Discussion This study is by far the most detailed available on the dietary habits of N. nigricollis and N. melanoleuca, two species that are known as generalist predators adapted to feed upon a wide variety of small vertebrates (e.g., see Luiselli and Angelici 2000). In this regard, the present study also fully confirms that these cobras may prey upon both endotherms and ectooterms, and upon terrestrial as well as arboreal and even aquatic organisms. The same has been also observed in other large African elapids, e.g., members of the genera Pseudohaje (Pauwels et al. 1999) and Dendroaspis (Luiselli et al. 2000b), and so is likely a general pattern for Afrotropical elapids. Nonetheless, the present study has revealed several other remarkable patterns of the foraging ecology of the two cobra species studied. To begin with, our study has demonstrated that intersexual differences in taxonomic prey composition were minor in

62 Can. J. Zool. Vol. 80, 2002 both species and in all the habitat types, which is consistent with the minor sexual size dimorphism observed. In fact, it has been demonstrated that intersexual dietary divergence in snakes is normally associated with pronounced size dimorphism between the sexes (e.g., see Shine 1986). Nonetheless, the analysis of body-size variation in cobras in relation to habitat type provides firm evidence of a particularly remarkable pattern in one of the two species. In fact, whereas in N. nigricollis the mean SVLs of the examined specimens did not vary among habitats, in N. melanoleuca there was a significant trend towards a reduction in mean body size from mature forest to suburbia, which is likely linked to the suboptimal adaptation of this species to strongly altered habitats (Luiselli and Angelici 2000). Unfortunately, no data on cobra sizes in other altered habitats are available in the literature, so any conclusion should be drawn with caution. The fact that N. melanoleuca attains smaller sizes in altered habitats than in forest habitats may depend on (i) a higher mortality rate of large specimens as a result of human activity (i.e., the snakes are generally killed before they attain a large size, i.e., old age), and (or) (ii) relative scarcity of preferred food resources in altered habitats. We have no data with which to test hypothesis i (although we are led to believe that it probably plays an important part), but we suggest that hypothesis ii should be accepted, also in the light of the better body condition exhibited by cobras captured in natural habitats. Moreover, our N. melanoleuca specimens from suburbia fed less frequently than those from the other two habitat types, although significantly more frequently than suburban N. nigricollis. In this case, however, we are led to believe that interspecific comparisons of feeding rates may not be useful because they may reflect intrinsic (i.e., species-specific) differences in metabolic rate, etc. Nevertheless, N. nigricollis is known to have lower proportions of individuals that had recently fed (i.e., lower feeding rates) than N. melanoleuca in every habitat type (see Luiselli and Angelici 2000; this paper). If we look at the interspecific differences in prey composition, it is evident that the two species fed, in part, on different species, although the respective dietary spectra were wide in all cases. Luiselli and Angelici (2000), who studied a smaller sample, suggested that N. nigricollis probably forages in drier microhabitats than N. melanoleuca, and that this is probably one of the major ecological differences between the two species. Our data confirm and further extend the above-mentioned hypothesis, especially with regard to cobras from suburbia and plantation forest mosaics. In addition, N. nigricollis reduced feeding rates during the peak of the dry season, whereas the same pattern was not found in N. melanoleuca (Luiselli 2001). However, as already explained in the Introduction, it is important to know whether the ecological success of N. nigricollis reflects the fact that it shows greater dietary plasticity than N. melanoleuca. In terms of prey size predator size relationships, it is now clear that the two species are very similar; they certainly do not differ in such a way as to support the above hypotheses. However, N. nigricollis tended to prey on slightly larger organisms (relative to its own body size) than N. melanoleuca, and this may help in sites where food-resource availability is limited (e.g., in some suburban habitats). In terms of diet flexibility, both species proved to be flexible predators indeed! Thus, it is unlikely that N. melanoleuca does not have the potential for efficiently colonizing the suburban habitats of southern Nigeria. So why is N. nigricollis more able to establish populations in suburban habitats? We think that the secret of its success is not its dietary flexibility but the consistency with which the juveniles prey upon a single prey type, i.e., lizards. In fact, agamid lizards are so abundant in nearly every altered habitat in Nigeria that they can be a virtually unlimited food resource for young N. nigricollis. Moreover, the fact that the adults forage frequently upon commensal mice and poultry also helps this species to colonize human-altered habitats, although it is likely that the same foraging strategy is used by adult N. melanoleuca. In any case, it is obvious that juvenile N. nigricollis prey mainly on lizards, and this could be taken as evidence supporting the above hypothesis that predation upon lizards is a foraging innovation which reflects behavioural flexibility. Certainly, if we consider the overall distribution of the two species (cf. Chippaux 1999), it is also possible that deforestation (with its impact on the microclimate, which becomes drier, parching the ground and causing many ponds and streams to dry up) may be the main factor causing the disappearance of N. melanoleuca, and that the colonization efficiency of N. nigricollis may then be favoured by the disappearance of a possible competitor. Acknowledgements We are indebted to several companies that supported our continued research in Nigeria. In particular we extend our thanks to IAM Oil Services Nigeria Ltd. (Port Harcourt), Prime Energy Resources Nigeria Ltd. (Port Harcourt), Agip-Petroli S.p.A. (Milan), Aquater S.p.A. (San Lorenzo in Campo), Technip, Kellogg Brown & Root, Japan Gas Company Corporation (T.S.K.J.) Nigeria Ltd. (Port Harcourt), Nigerian Agip Oil Company Ltd. (Port Harcourt), Nigerian Agip Exploration Ltd. (Lagos), Ecosystem s.r.l. (Bari), Amertex Oil and Gas Ltd. (Lagos), Italian Foundation of Vertebrate Zoology (Rome), Chelonian Research Foundation (Lunenburg), and Demetra S.p.A. (Fano and Rome). E. Politano, I.F. Barieenee, B. Ekeke, L.D. Otonye, D. Capizzi, A. Pazienti, G. Paoloni, L. Rugiero, J.S. Ekanem, B. Egbide, L. Ude, L. Bikikoro, B. Bolton, C. Effah, M.A. Inyang, S. Kalio, A. Sigismondi, and Z. Tooze spent much pleasant time with us in the field and gave some helpful observations. M. Capula, R.W. Henderson, N.J. Scott, and S. Akele offered helpful critical comments on a previous version of the manuscript, and R. Shine responded to all our queries. References Altmann, S.A. 1998. Foraging for survival: yearling baboons in Africa. Chicago University Press, Chicago. Brandon, R.N. 1995. Adaptation and environment, Princeton University Press, Princeton, N.J. Brooke, R.K., Lockwood, J.L., and Moulton, M.P. 1995. Patterns of success in passeriform bird introductions on Saint Helena. Oecologia, 103: 337 342. Capizzi, D. 2000. Diet shifts of the tawny owl Strix aluco in central and northern Italy. Ital. J. Zool. 67: 73 79. Case, T.J. 1996. Global patterns in the establishment and distribution of exotic birds. Biol. Conserv. 78: 69 96.

Luiselli et al. 63 Chippaux, J.-P. 1999. Les serpents de l Afrique Occidentale et Centrale. IRD Editions, Paris. De Montclos, M.-A. 1994. Le Nigéria. Kurthala, Paris. Ehlrich, P. 1989. Attributes of invaders and the invading processes: vertebrates. In Biological invasions: a global perspective. Edited by J.A. Drake, H.A. Mooney, F. di Castri, R.H. Groves, F.J. Kruger, M. Rejmanek, and M. Williamson. John Wiley and Sons, London. pp. 315 328. Losos, J.B., and Spiller, D.A. 1999. Differential colonization success and asymmetrical interactions between two lizard species. Ecology, 80: 252 258. Lovari, S., Renzoni, A., and Fondi, R. 1976. The predatory habits of the barn owl (Tyto alba) in relation to the vegetation cover. Boll. Zool. 43: 173 191. Luiselli, L. 2001. The ghost of a recent invasion in the reduced feeding rates of spitting cobras during the dry season in a rainforest region of tropical Africa? Acta Oecol. 22: 311 314. Luiselli, L., and Angelici, F.M. 2000. Ecological relationships in two Afrotropical cobra species (Naja melanoleuca and Naja nigricollis). Can. J. Zool. 78: 191 198. Luiselli, L., Politano, E., and Angelici, F.M. 2000a. Ecological correlates of the distribution of terrestrial and freshwater chelonians in the Niger Delta, Nigeria: a biodiversity assessment with conservation implications. Rev. Ecol. Terre Vie, 55: 3 23. Luiselli, L., Angelici, F.M., and Akani, G.C. 2000b. Large elapids and arboreality: the ecology of Jameson s green mamba (Dendroaspis jamesoni) in an Afrotropical forested region. Contrib. Zool. 69: 147 155. Moulton, M.P., and Pimm, S.L. 1983. The introduced Hawaiian avifauna: biogeographic evidence for competition. Am. Nat. 121: 669 690. Pauwels, O.S.G., David, P., and Lenglet, G. 1999. Pseudohaje goldii (Boulenger, 1895): Goldie s tree cobra diet. Afr. Herpetol. News, 30: 32 33. Petren, K., and Case, T.J. 1998. Habitat structure determines competition intensity and invasion success in gecko lizards. Proc. Natl. Acad. Sci. U.S.A. 95: 11 739 11 744. Pimm, S.L. 1991. The balance of nature? University of Chicago Press, Chicago. Politano, E. 1997. A study of the fauna (Amphibia, Reptilia, Aves, Mammalia) of the Niger Delta, and environmental assessment of the works of two gas pipelines. E.N.I. Press, Milan, Italy. Shine, R. 1986. Sexual differences in morphology and niche utilization in an aquatic snake, Acrochordus arafurae. Oecologia, 69: 260 267. White, F. 1983. The vegetation of Africa UNESCO, Paris. Williamson, M. 1996. Biological invasions. Chapman and Hall, London.