Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Niche overlap between marsupial and eutherian carnivores: does competition threaten the endangered spotted-tailed quoll?
A. S. Glen,
School of Biological Sciences, A08, University of Sydney, NSW 2006, Australia
*Correspondence and present address: Department of Environment and Conservation and Invasive Animals CRC, Dwellingup Research Centre, Banksiadale Road, Dwellingup, WA 6213, Australia. E-mail: firstname.lastname@example.org
*Correspondence and present address: Department of Environment and Conservation and Invasive Animals CRC, Dwellingup Research Centre, Banksiadale Road, Dwellingup, WA 6213, Australia. E-mail: email@example.com
1The significance of top-down regulation by carnivores is receiving increasing global recognition. As a consequence, key objectives in many programmes that seek to maintain ecosystem function now include conserving carnivores and understanding their interactions. This study examined overlap in resource use (space and diet) of introduced eutherian carnivores and an endangered marsupial carnivore, the spotted-tailed quoll Dasyurus maculatus, in eastern Australia. We also investigated mechanisms of niche partitioning and evidence for interspecific aggression.
2Dietary overlap between quolls, red foxes Vulpes vulpes and wild dogs Canis lupus ssp. was assessed by analysis of scats. Trapping, radio-tracking and direct observations were used to quantify spatial overlap between quolls, foxes, wild dogs and feral cats Felis catus.
3Dietary overlap among the carnivores was extensive. Medium-sized mammals were the most important prey for all three predators, indicating potential for exploitative interactions. However, hunting of different size classes of secondary prey and consumption by quolls of more arboreal prey than their counterparts may assist coexistence. Remains of quoll were found in two dog scats, and cat hair in another, possibly indicating intraguild predation.
4We observed extensive spatial overlap between quolls and eutherian carnivores. However, we inferred from dietary data that quolls foraged primarily in forested habitat, while canids foraged mainly in cleared habitat.
5Synthesis and applications. Our results indicate strong potential for competition between spotted-tailed quolls and eutherian carnivores, and thus a situation where control of introduced predators may be desirable, not only for the conservation of prey species but also for the protection of native carnivores. Concern over potential non-target mortality of quolls has hindered efforts to control foxes in eastern Australia using poison baits. We contend that, rather than harming quoll populations, baiting for foxes should aid the conservation of quolls and should be implemented in areas of sympatry where fox numbers are high.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Competition between mammalian predators often has a strong influence on their abundance and distribution. For example, African wild dogs Lycaon pictus (Temminck) are more abundant in the absence of lions Panthera leo (L.) and spotted hyaenas Crocuta crocuta (Erxleben) (Creel 2001; Creel, Spong & Creel 2001) and red foxes Vulpes vulpes (L.) have extirpated populations of arctic foxes Alopex lagopus (L.) through competitive interference (Hersteinsson et al. 1989; Kaikusalo & Angerbjörn 1995). In eastern Australia, the spotted-tailed quoll Dasyurus maculatus (Kerr), an endangered marsupial carnivore, is thought to be threatened by competition from eutherian carnivores such as the red fox, feral cat Felis catus (L.) and wild dog Canis lupus (L.) ssp. (Glen & Dickman 2005).
Dominant competitors often exclude subordinates from some microhabitats (Dickman 1988), or even from larger areas, so that little or no spatial overlap occurs between the species (Harrison, Bissonette & Sherburne 1989). If competition occurs between spotted-tailed quolls and eutherian predators, such inverse relationships might be expected between the abundances and distributions of these species. Indeed, the highest abundances of spotted-tailed quolls occur in areas where foxes are rare or absent (Catling & Burt 1994, 1997). Despite their intuitive appeal, however, associations such as these could also reflect differences in habitat preference or distributions of prey species, and therefore present only weak evidence for competition. Evidence is more convincing if species overlap substantially in resource use, while conclusive evidence is usually obtained only from appropriate manipulative experiments (MacNally 1983; Wiens 1989).
The aims of this study were to (i) quantify the degree of overlap in resource use between sympatric quolls, foxes, feral cats and wild dogs; (ii) investigate possible mechanisms of niche partitioning (dietary or spatial); and (iii) examine any evidence of interspecific aggression or killing. We used this information to assess the likelihood of competition and to make recommendations for future management.
The study was conducted in the Marengo and Chaelundi State Forests, north-eastern New South Wales, Australia (centre point 30°07′ S, 152°23′ E). The altitude ranges from 900 to 1300 m, and the mean annual rainfall is 1600–2000 mm. The area is characterized by open, dry sclerophyll forest with a grassy understorey, although temperate rainforest typically covers gullies and creek lines. In the centre of the study site is a cleared area of private leasehold land (c. 4 km2), known as the Marengo Plain, on which cattle are grazed. The study area was not subject to predator control and hence no disruption to patterns of resource use overlap were expected.
We investigated the diets of spotted-tailed quolls, foxes and wild dogs by identifying prey remains in scats. A detailed description of the dietary analysis technique is provided in Glen & Dickman (2006a). We identified scats of each species based on shape, size and odour. Identification was confirmed in some cases by the detection of small numbers of grooming hairs during microscopic analysis. However, as scats can be misidentified even by an experienced observer (Davison et al. 2002), we excluded scats of uncertain origin from analyses. We collected scats from trapped animals and five 1-km road transects that were cleared on a monthly basis between January 2003 and February 2004. Opportunistic collection of scats continued within the same area until October 2004.
For each predator species, we plotted the cumulative diversity of prey items against sample size to determine whether sufficient scats had been analysed to describe the diet accurately. Diversity was calculated using the Brillouin index. This is calculated according to the equation:
H = ln N!–Σ ln ni!/N
where H = diversity, N = total number of individual prey recorded and ni= number of individual prey items in the ith category (Brillouin 1956). Based on the frequency of occurrence of each prey category, we calculated dietary overlap between each pair of species using Pianka's index, which ranges from zero (signifying no overlap) to one (complete overlap) (Pianka 1973). This index was chosen to allow direct comparisons of the degree of overlap in similar studies of carnivores elsewhere (Fedriani et al. 2000; Ray & Sunquist 2001; Jácomo, Silveira & Diniz-Filho 2004). Because dietary niche separation commonly occurs through a variety of mechanisms, we conducted separate tests for separation according to (i) prey size or taxa and (ii) arboreal or terrestrial prey. We investigated differences in the frequency of occurrence of prey sizes or taxa between predators by pairwise comparison using chi-squared contingency tests with sequential Bonferroni adjustments (Quinn & Keough 2002). For this purpose, food types were divided into seven categories: small mammals (1–499 g), medium-sized mammals (500–6999 g), large mammals (7 kg or more), insects, birds, reptiles and vegetation. Mammalian prey were assigned to size categories according to the maximum weights listed by Menkhorst & Knight (2001). We also made interspecific comparisons of the frequency of occurrence of arboreal possums and gliders (the staple prey of quolls) and rabbits Oryctolagus cuniculus (L.) (the staple prey of foxes and wild dogs). Differences in the diets of quolls, foxes and wild dogs in terms of volume were investigated using non-metric multidimensional scaling (MDS) and global one-way analysis of similarities (anosim), based on a Bray–Curtis similarity matrix (Clarke & Warwick 1994). We also used unpublished data on the contents of 92 wild dog scats collected in and around our study area in 1984 by R. Harden & J. Robertshaw, to make inferences regarding the possibility of intraguild predation.
We used radio-telemetry, trapping and opportunistic sightings to investigate the use of space by predators. Quolls were captured between January 2003 and October 2004 in wire cage traps (30 × 30 × 60 cm; Mascot Wireworks, Sydney, Australia) baited with chicken wings. For foxes and feral cats, we used Victor Soft Catch leg-hold traps (Woodstream Corporation, Lititz, Pennsylvania) baited with a range of olfactory lures, including synthetic fermented egg, berry flavour essences and meat. Leg-hold traps were set on and around the Marengo Plain for about 600 trap nights between October 2003 and October 2004.
We fitted radio-collars (Faunatech, Bairnsdale, Australia; 150–151 MHz) to eight quolls (four male, four female), two feral cats (one male, one female) and one male fox. Animals were located using triangulation while they were active, or by approaching them when inactive in dens or shelter sites. Where possible, animals were located once each day and again each night (for more details see Glen & Dickman 2006b). The radio-collars were equipped with a mortality function; the pulse rate would double from 50 to 100 pulses per minute if no movement occurred for 12 h, indicating that the animal had died or shed its collar. Spatial overlap between radio-collared individuals was assessed using minimum convex polygon (MCP) home ranges estimated for animals that were radio-tracked over corresponding time periods. Home range estimates were derived using the animal movement extension to ArcView (Hooge & Eichenlaub 1997). Capture locations were included in the analysis of home ranges.
The locations of all predator sightings were recorded using a global positioning system (GPS). These data were used to confirm spatial overlap between species.
Scat collection yielded 424 quoll, 207 fox, 73 dog and three cat scats. The low number of cat scats, perhaps because of the species’ habit of burying its faeces, meant that no meaningful analysis of cat diet could be made. The dietary study was therefore restricted to quolls, foxes and wild dogs. Extensive overlap occurred between the diets of these three species.
The cumulative diversity (Hk) of foods in the diet of each species approached an asymptote at a sample size well below the number of scats analysed (Fig. 1), indicating that the sample sizes were adequate. A total of 43 food types was identified in the scats (Table 1). Twenty-two food types (51%) were common to the diets of all three predators. In terms of frequency of occurrence, medium-sized mammals dominated the diets of all species, in particular rabbits, red-necked pademelons Thylogale thetis (Lesson) and bandicoots Perameles nasuta (Geoffroy) and Isoodon macrourus (Gould). Pairwise comparisons indicated extensive overlap in the diets of all three predators. Pianka's index of dietary overlap between quolls and foxes was 0·712, while overlap between quolls and dogs was 0·657. Foxes and dogs showed the greatest similarity in diet, with an index value of 0·943.
Table 1. Frequency of occurrence (%) of prey taxa in the diets of spotted-tailed quolls (n = 424), foxes (n = 207) and wild dogs (n = 73) in Marengo and Chaelundi State Forests (modified from Glen & Dickman 2006a)
May include plant material ingested incidentally with prey.
Antechinus swainsonii, dusky antechinus
Melomys spp., grassland and fawn-footed melomys
Rattus lutreolus, swamp rat
Rattus rattus, black rat
Rattus fuscipes, bush rat
Mus domesticus, house mouse
Muridae, unidentified rodents
Chiroptera, unidentified bat
Petaurus breviceps, sugar glider
Petauroides volans, greater glider
Oryctolagus cuniculus, rabbit
Peramelidae, long-nosed and northern brown bandicoots
Thylogale thetis, red-necked pademelon
Pseudocheirus peregrinus, ring-tailed possum
Trichosurus spp., common and mountain brushtail possums
Potorous tridactylus, long-nosed potoroo
Tachyglossus aculeatus, echidna
Macropus parma, parma wallaby
Petaurus australis, yellow-bellied glider
Dasyurus maculatus, spotted-tailed quoll
Aepyprymnus rufescens, rufous bettong
Wallabia bicolor, swamp wallaby
Bos taurus, cattle
Macropus giganteus, eastern grey kangaroo
Macropus spp., eastern grey kangaroo or red-necked wallaby
Despite the high degree of similarity (Fig. 2), there were significant differences between predators regarding the frequency of occurrence of some prey categories in the diets. Foxes ate small mammals more often than quolls (χ2 = 44·6, d.f. = 1, P < 0·001) and dogs (χ2 = 10·3, d.f. = 1, P= 0·001) did. Conversely, dogs ate large mammals more often than foxes (χ2 = 34·0, d.f. = 1, P < 0·001) and quolls (χ2 = 20·9, d.f. = 1, P < 0·001) did. Insects were consumed more often by quolls than by foxes (χ2 = 9·1, d.f. = 1, P= 0·003) and dogs (χ2 = 12·2, d.f. = 1, P < 0·001). Foxes ate birds more often than quolls did (χ2 = 15·9, d.f. = 1, P < 0·001). Both wild dogs (χ2 = 49·2, d.f. = 1, P < 0·001) and foxes (χ2 = 42·5, d.f. = 1, P < 0·001) consumed vegetation more often than quolls did.
Differences also occurred in terms of the most commonly eaten prey. Arboreal possums and gliders were eaten more frequently by quolls than foxes (χ2 = 17·4, d.f. = 1, P < 0·001) and wild dogs (χ2 = 16·6, d.f. = 1, P < 0·001). Conversely, rabbits were consumed more often by foxes (χ2 = 63·6, d.f. = 1, P < 0·001) and dogs (χ2 = 47·4, d.f. = 1, P < 0·001) than by quolls.
Multidimensional scaling confirmed the high degree of dietary overlap between quolls, foxes and dogs in terms of volume of prey types in the diet, with samples from all three species overlapping extensively in ordination space (Fig. 3). In terms of the relative volumes of prey categories, global one-way anosim revealed no significant difference in the diets of quolls, foxes and dogs (global R= 0·027, P= 0·113). However, pairwise comparison between the species showed a small but significant difference between the diets of quolls and foxes (R = 0·04, P= 0·02).
There was no evidence of intraguild predation in our data set but analyses of 92 wild dog scats by R. Harden & J. Robertshaw (unpublished data) revealed that dogs had consumed smaller predators. Spotted-tailed quoll hair occurred in two scats and cat hair in one scat.
spatial resource use
Radio-collars fitted to animals remained operative for 1–6 months. The mean number of independent locations recorded for each animal was 39 (range 13–74). The mean MCP home range estimate for female quolls was 133 ha, while the mean estimate for males was 363 ha (Glen & Dickman 2006b). The MCP home range of the male feral cat was estimated at 432 ha, the female cat at 83 ha, and the fox at 227 ha. As only two home ranges (those of one male and one female quoll) were considered to be adequately defined (i.e. the last five location records increased the MCP area estimate by less than 5%), the remainder were likely to be underestimates. Despite this, the home ranges of quolls clearly overlapped extensively with those of eutherian predators (Fig. 4). This was confirmed further by the locations of predator sightings. Three fox, nine cat and four dog sightings were recorded within the known home ranges of quolls. With two exceptions, the individuals with overlapping home ranges were known (either from radio-tracking or trapping data) to occupy the areas of overlap at the same time. There was no mortality of radio-collared animals and therefore no direct evidence of interspecific killing.
There was some evidence of the use of similar den sites by quolls and foxes. As well as 38 quoll dens described by Glen & Dickman (2006b), three dens belonging to the same radio-collared fox were also located. All three dens were in crevices among rock jumbles, which is a type of den structure commonly used by spotted-tailed quolls. Unlike the quolls, the radio-collared fox did not change dens frequently and was using the same den on all but two of the 26 occasions when he was located during the day. The radio-collared cats frequently fled their diurnal rest sites when approached and could not be located precisely. However, on one occasion, the radio-collared male cat was found resting in a dense pile of fallen timber.
Our results demonstrate extensive overlap in resource use between all four species of mammalian carnivores in Marengo and Chaelundi State Forests. As outlined by MacNally (1983) and Wiens (1989), extensive overlap in resource use helps to corroborate the relatively weak evidence for competition that comes from inverse relationships in abundance and distribution. Taken together, these two forms of evidence present a stronger (although still inconclusive) case for the existence of competition. Is the overlap in resource use measured here high in comparison with other carnivore guilds and, if so, what facilitates the coexistence of quolls and eutherian predators? We will explore these two questions in turn, first in terms of dietary overlap, then in terms of spatial resources. The potential for competition is then discussed.
The diets of spotted-tailed quolls, foxes and wild dogs in this study were similar, not only in terms of the range of prey types consumed but also their frequency of occurrence and relative volumes. Differences in methodology make direct comparison with other dietary studies difficult. However, a number of previous studies have used Pianka's index to estimate dietary overlap within mammalian carnivore guilds. For example, in a study of eight sympatric carnivore species in central Africa (Ray & Sunquist 2001), 17 of 21 pairwise comparisons yielded a Pianka's index lower than those calculated in the present study. Among the three species with the highest dietary overlap, temporal and vertical niche partitioning occurred (Ray & Sunquist 2001). Similarly, within a carnivore guild in California, Fedriani et al. (2000) calculated Pianka's indices ranging from 0·52 to 0·79. Once again, avoidance was apparent between subordinate and dominant competitors. Conversely, in a dietary study of three sympatric canids in Brazil, Jácomo, Silveira & Diniz-Filho (2004) calculated Pianka's indices ranging from 0·044 to 0·498, indicating that dietary overlap within that guild was much lower than that found in the present study. Based on these limited comparisons, it appears that dietary overlap between many sympatric carnivore species is lower than we observed. Where comparable levels of overlap have been observed, temporal or spatial separation between species has typically been found. As discussed below, our data suggest similar mechanisms of separation.
The similarity found here between the diets of foxes and wild dogs is consistent with the findings of Mitchell & Banks (2005), who reported a Pianka's index of 0·91 in the central tablelands of New South Wales. However, a meta-analysis of previous dietary studies from similar areas suggested a much lower Pianka's index of 0·69 (Mitchell & Banks 2005). Similarly, Glen, Fay & Dickman (2006) reported a Pianka's index of 0·697 between the diets of foxes and wild dogs in the Northern Rivers region of New South Wales. In both prior studies, as in the present one, foxes ate small prey, and dogs large prey, more often than their counterparts (Mitchell & Banks 2005; Glen, Fay & Dickman 2006).
Despite high similarity in the diets of predators in this study, competition may still not occur unless resources are limiting (Schoener 1983). However, there were several significant differences between species that may facilitate their coexistence even if resources are limiting. A high degree of vertical niche partitioning was found between the diets of spotted-tailed quolls and canids. Greater gliders Petauroides volans (Kerr) in particular were eaten very frequently by quolls and infrequently by foxes and wild dogs. Greater gliders are arboreal (Kehl & Borsboom 1984) and, as such, are probably relatively inaccessible to canids, which have limited climbing ability. Spotted-tailed quolls, however, are adept climbers, and have been observed climbing trees to capture gliders while they are inactive in nest hollows (Belcher, Nelson & Darrant 2007). Thus quolls probably had almost exclusive access to greater gliders, which were their staple prey.
In addition to the evident vertical partitioning of prey taxa, each of the canids appeared to exploit a different size class of secondary prey. While medium-sized mammals were the staple prey of all three predators, foxes consumed small mammals, and dogs large mammals, significantly more often than their counterparts. Conversely, quolls ate more insects than foxes or dogs. However, this was largely because of a sharp peak in the occurrence of insects during the summer; insects were rare or absent in the diet at other times (Glen & Dickman 2006a).
Both canids consumed vegetation more often than quolls did. This may reflect the more omnivorous nature of the canids compared with the spotted-tailed quoll. However, because it was not possible to determine how much of the vegetation in scats had been consumed incidentally with prey, this result must be interpreted cautiously.
Although the diet of cats was not sampled here, previous studies reviewed by Dickman (1996) suggest considerable similarity with the diet of the spotted-tailed quoll. Cats were common in the study area (A. Glen, unpublished data; Forests NSW and Department of Environment and Conservation, unpublished data) and may also have the potential to compete with quolls for food.
Radio-tracking showed extensive spatial overlap between spotted-tailed quolls, foxes and feral cats in the study area. This was confirmed further by the locations of predator sightings, which also indicated spatial overlap between quolls and wild dogs. In some systems, coexistence between predator species is characterized by interspecific territoriality, with subordinate species mainly utilizing areas outside the home ranges of dominant predators (Harrison, Bissonette & Sherburne 1989; Johnson & Franklin 1994; Gosselink et al. 2003). Spatial avoidance at such a broad scale did not occur between spotted-tailed quolls and the eutherian predators in our study. For example, the home range of one female quoll was contained almost entirely within that of a large male feral cat (Fig. 4) while three of the four radio-collared male quolls had home ranges overlapping that of the radio-collared fox. As actual home ranges were almost certainly larger than those estimated, spatial overlaps between species were probably also greater. However, spatial overlap does not preclude the possibility that animals avoid direct encounters. Avoidance may still occur at a finer spatial scale (Molsher 1999), animals may be active at different times of day (Ray & Sunquist 2001) and they may partition the habitat vertically while still using the same two-dimensional area (Jones & Barmuta 2000; Ray & Sunquist 2001). Animals with overlapping home ranges may also forage preferentially in different habitats (Loveridge & Macdonald 2003). There was some evidence for this in the present study as quolls relied heavily on forest-dwelling prey such as greater gliders, while foxes and dogs more frequently ate rabbits, which inhabited the cleared area of the Marengo Plain.
Evidence from previous studies confirms that northern D. hallucatus (Oakwood 2000) and spotted-tailed (Körtner et al. 2004; Körtner & Watson 2005; Körtner 2007) quolls are frequently killed by other predators such as feral cats, foxes and wild dogs. Similarly, wild dogs in and around our study area consumed spotted-tailed quoll and cat (R. Harden & J. Robertshaw, unpublished data). This may indicate intraguild predation by dogs or that these animals were scavenged. Quolls killed by larger predators may not always be eaten (Körtner & Watson 2005; Körtner 2007). Therefore, although such interspecific killing is known to occur, its incidence may be underestimated by dietary studies.
Although there was no mortality among the radio-collared animals in the present study, the small number of collared animals means that interspecific killing was unlikely to be detected. The possibility cannot therefore be discounted. The frequent den shifting of quolls (Oakwood 1997; Glen & Dickman 2006b) may serve to minimize any risk of predation. Similarly, greater gliders alternate between several nest hollows in order to avoid predators (Kehl & Borsboom 1984).
Despite the apparent vertical niche partitioning of diet, quolls did not appear to favour arboreal habitats in order to reduce the risk of encounter with canids. Spool-and-line tracking revealed that quolls were usually active at or close to ground level (Glen & Dickman 2006b). However, it remains possible that the climbing ability of quolls provides a means of escape in direct encounters with other predators, thereby reducing the severity of interference competition. For example, grey foxes climb trees to escape aggression from coyotes Canis latrans (Cypher 1993) and Antechinus agilis climbs to avoid direct encounters with the dominant Antechinus swainsonii (Dickman 1991).
conclusions and management recommendations
The high degree of resource overlap measured here indicates the potential for exploitation competition, not only between quolls and eutherian predators, but also between the eutherian predators themselves. Coexistence in this predator assemblage is likely to be facilitated by vertical niche partitioning of the diets of quolls and canids, and the exploitation by foxes and dogs of different size classes of secondary prey. If interspecific aggression occurs, this may to some extent be alleviated by the ability of quolls to climb trees to escape. The apparently low density of foxes in the study area (Forests NSW and Department of Environment and Conservation, unpublished data) may also have contributed to the high abundance of quolls. Increased numbers of introduced predators could potentially exert much greater competitive pressure on spotted-tailed quolls, through both interference and exploitation of common resources.
Although areas exist (including the present study area) where sympatric spotted-tailed quolls, wild dogs and feral cats apparently all occur in abundance, the same does not appear to be true of quolls and foxes. Indeed, following an extensive search for sites at which to conduct the present study, we were unable to locate any areas where both foxes and quolls were numerous. In western Australia, widespread fox control has been associated with the dramatic recovery of western quolls Dasyurus geoffroii (Gould) (Morris et al. 2003). In eastern Australia, concern over the potential non-target mortality of spotted-tailed quolls has hindered efforts to control foxes using 1080 poison baits. However, recent research, reviewed by Glen, Gentle & Dickman (2007), suggests that quoll populations face little risk of non-target poisoning. We contend that widespread fox control, rather than harming populations of spotted-tailed quolls, would be of great benefit in their recovery, and that 1080 baiting should be implemented in areas of sympatry where fox numbers are high.
Further research is also required to clarify the interactions between carnivores in Australia. Unlike carnivore guilds in many parts of the world that evolved in sympatry, foxes and cats have coexisted with native Australian carnivores for only a short time. This may hinder direct comparison of our results with those from other studies but also allows the unparalleled opportunity to study competition without the confounding influence of coevolution (Clode & Macdonald 1995; Blackwell & Linklater 2003). Networks of direct and indirect interactions probably exist within suites of most sympatric predators (Glen & Dickman 2005) and manipulation experiments are needed to tease out the pairwise relationships between the species.
Funding was provided by the Pest Animal Control CRC, NSW Department of Environment and Conservation (DEC), Australian Geographic Society, Australian Academy of Science, Wildlife Preservation Society of Australia, Foundation for National Parks and Wildlife and the Royal Zoological Society of NSW. All procedures comply with Australian law and were approved by the University of Sydney Animal Ethics Committee (Approval No. L04/7-2002/2/3589). Research was licensed by DEC (licence no. S10566) and Forests NSW (permit no. 14990). Sincere thanks to P. Meek, B. Tolhurst, J. Haydock, J. Bertram, A. Lloyd and D. Everson for assistance and accommodation in the field, and to R. Harden, G. Körtner and O. Albanil, who shared equipment, ideas and advice. We are grateful to P. Fleming for advice and to R. Harden and J. Robertshaw for providing unpublished data that were of great value. Sincere thanks also to M. Letnic and M. Crowther for comments that helped greatly to improve the quality of the manuscript.