INNATE PREDATOR-RECOGNITION IN AUSTRALIAN BRUSH-TURKEY (ALECTURA LATHAMI, MEGAPODIIDAE) HATCHLINGS by A. GÖTH1) (Australian School of Environmental Studies, Grif th University, Brisbane, Nathan 4111 QLD, Australia, e-mail A.Goeth@mailbox.gu.edu.au) (Acc. 1-XI-2000) Summary Hatchlings of the Australian brush-turkey, Alectura lathami, should respond to predators innately because they hatch independently of nest-mates, have no contact with parents, and initially live solitarily. Their response to predators was tested in a large outdoor aviary set in natural rainforest habitat. Two living predators, a cat and a dog, as well as a moving rubber snake and raptor silhouette were presented to observe whether different predators evoked different innate responses. Controls consisted of cardboard boxes of equal coloration, shape and dimensions. Ten chicks were tested per stimulus type, and their response measured as latency to the rst step and proportion of time spent performing different behaviours, during presentation of the stimuli and thereafter. While the snake evoked mainly running and this was obvious only during the test, the three other stimuli also led to a difference in behaviour 1) I thank the Dr. Otto Röhm Gedächnisstiftung, the Australian Geographic Society, J. & Ch. Haas, U. Wehrhahn and the World Pheasant Association, including WPA Germany, for nancial assistance. E. Curio, through his Bird Research and Conservation Foundation, helped with the nancial administration. E. Curio, D. Jones, I. McLean, H. Proctor and U. Vogel gave helpful suggestions during the preparation phase for this study. J. Brooks and U. Vogel assisted with the experiments. Permission to build an aviary in Mary Cairncross Scenic Reserve and to dig out brush-turkey eggs was given by the Caloundra City Council. D. Jones and I. McLean considerably improved earlier drafts of this paper. This study was undertaken with a Scienti c Purposes Permit from the Queensland Department of Environment and approval from the Grif th University Ethics Committee for Animal Experimentation. An Overseas Postgraduate Research Award from the Australian Government, a Grif th University Postgraduate Research Scholarship and a research allowance from the Australian School of Environmental Studies, Grif th University, supported my study. c ° Koninklijke Brill NV, Leiden, 2001 Behaviour 138, 117-136 118 A. GÖTH after presentation. The raptor and cat evoked more crouching than other stimuli and the dog more running. Latency to the rst step was higher in the raptor tests than during others. However, there was no difference in response between the stimuli and controls, suggesting that the releasing mechanism for evoking a response is likely to be size, dimensions, height and/or relative speed. Hatchlings were also presented with an acoustical stimulus, alarm calls of songbirds; its control was white noise. They responded to this by being more vigilant than in other tests, and, as with the snake, this response was only obvious during the test. In contrast to the optical stimuli, chicks did not respond to the control for the acoustical stimulus, indicating that megapode chicks, which have no parents to warn them, possess an innate response to alarm calls of songbirds instead. The results of this study also suggest that a lack of predator recognition should be of little concern in the translocation of endangered megapode species, even when chicks have to deal with introduced predators, and that other factors such as the availability of cover should be given greater attention. Introduction Predator-recognitio n has a strong innate component in many birds (e.g. Cu- rio, 1975). To date, this has mainly been demonstrated in adult or sub-adult birds that were predator-naive, either because they occurred allopatrically from some species of predators (e.g. on islands), or because they had been hand-raised (summarised in Curio, 1993). Young birds have been studied less often and their response to predators often described in very general terms (e.g. ‘innate fear response to a wide variety of stimuli’; Buitron, 1983; Dow- ell, 1986). Schaller & Emlen (1962) found that ‘avoidance behaviour’ was not exhibited by hatchlings of eight precocial bird species but appeared grad- ually during the rst week of post-natal life. Most other researchers raised young birds for at least a week before testing them with typically a single stimulus, such as ying hawk and goose models (Krätzig, 1939; Tinbergen, 1963; Mueller & Parker, 1980), eye-like shapes (Jones, 1980) or human in- truders (Gallup et al., 1972). Other studies of predator-recognitio n in bird hatchlings are rare, presumably because they are logistically dif cult to con- duct. Bird hatchlings normally behave naturally only when kept with oth- ers, and it is dif cult to distinguish between responses directly related to the predator and responses in uenced by interactions with parents or sib- lings. Thus most studies on innate predator-recognitio n have been on sh and reptiles which often live independently from the moment of hatching (e.g. Brown & Warburton, 1997; Burger, 1998). Chicks of the Australian brush-turkey, Alectura lathami, and all other megapodes (family Megapodiidae) share several signi cant features with PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 119 sh and reptiles: they are highly precocial and form no bonds with parents or siblings, which makes them ideal subjects for a study on the existence of innate predator recognition in birds. They hatch underground where the egg had been incubated by external heat sources and, having emerged at the surface, lead a solitary life (for details see Jones et al., 1995). Without adults or parents to warn or protect them, they must recognise enemies, and respond to them, alone. Such responses can be termed innate because the hatchlings have no chance to experience a predator individually prior to hatching (Curio, 1993). But do chicks show the same type of response to any moving object or does their response vary with the kind of enemy, e.g. is it more ne-scaled? Fine-scaled predator-recognitio n has been found in some bird species, but is only known for adults (Hartzler, 1974; Curio, 1975; Buitron, 1983; Walters, 1990). This study investigates the possibility of a ne-scaled innate response in bird hatchlings. It aims at determining whether brush-turkey hatchlings respond differently to stimuli of different size and shape, a live cat and dog, a ying raptor model and a rubber snake moved through their aviary, as well as to their controls (cardboard boxes of equal shape, coloration and dimensions). Additionally, it aims at testing whether these chicks also respond to alarm calls of songbirds that live in the same habitat. Auditory cues are often important in predator-prey interactions (Conover & Perito, 1981). Since brush-turkey parents do not utter any alarm calls, I tested the hypotheses that chicks should respond to alarm calls of other birds instead. In a previous study, Wong (1999a) presented brush-turkey hatchlings in an indoor aviary with a live cat (in a box with window), a replica snake (not moving), video images of various predators, a live guinea pig and a toy rabbit. She could not detect any response to these stimuli, and it appeared likely that the performance of natural responses was confounded by the arti cial environment in which stimuli were presented — a problem in many studies of animal behaviour (Murphy, 1978). The present study was therefore conducted in a quasi-natural environment, a large outdoor aviary in the rainforest, and some of the stimuli were live animals that moved through the aviary. Studies of predator recognition in hatchlings also have implications for the conservation of endangered megapodes. Of the 22 species described by Jones et al. (1995), nine are listed as vulnerable and three as endangered (Dekker, 1999). Conservation plans typically do not consider the needs of 120 A. GÖTH the chicks (Dekker & McGowan, 1995; Dekker et al., in press), primarily because nothing is known about their behaviour in the wild (Jones, 1999). Information on the predator-recognitio n ability of megapode chicks has implications for management plans that deal with the protection and re- introductio n of endangered species. A re-introductio n of the endangered malleefowl, Leipoa ocellata, for example, was unsuccessful mainly because introduced foxes killed most of the released hatchlings (Priddel & Wheeler, 1994). It was not known whether this was due to a lack of response in the chicks, their physical condition when released, the choice of the release site, or other factors. Methods Study site and setting All trials took place in a quasi-natural setting, in a large (10 £ 8 m, 2.5-3.5 m height) outdoor aviary built of transparent shadecloth around the existing vegetation in a subtropical rainforest (Mary Cairncross Scenic Park, 55 ha, 120 km north of Brisbane, Queensland, Australia, 38± S, 88± E). Trees and logs provided natural roosting places; food and water were available ad libitum. All ground vegetation that would have disturbed the observations was cleared, except for two patches of thicket. The chick’s behaviour was observed from a blind incorporated into the walls of the aviary and tests were conducted between September and December 1999. Type of models The following ve stimuli were used: (1) a live dog (50 cm high, grey and black Australian blue cattle dog); (2) a live cat (20 cm high, grey); (3) a rubber snake (180 cm long, 4 cm in diameter, resembling a red-bellied black snake, Pseudechis porphyriacus); (4) a plywood silhouette of a raptor known to prey on brush-turkey chicks (grey goshawk, Accipiter novaehollandiae; wingspan 47 cm, head to tail length 24 cm); and (5) taped alarm calls of songbirds (yellow-throated scrubwrens, Sericornis citreogularis, directed against ground predators and recorded where brush-turkeys occur naturally). The control object for the four optical stimuli were cardboard objects that had an equal visible area, same coloration and similar shape (square for cat, dog and raptor, cylindrical for snake), while the control for the alarm calls was white noise. Throughout the text, the terms ‘stimulus’ and ‘control’ are used, and the word ‘model’ is used to refer to the combination of one particular stimulus and its control. Presentation of the models The type of model presented to a chick on a given day was chosen randomly. The blind was connected with the aviary via a door (40 £ 80 cm) closed with a curtain. Prior to the experimental period, the cat and dog had been trained to walk from the blind through the aviary and back using food rewards. During experiments, an assistant controlled their PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 121 movement using a loop rope and lead system. This system was also used to move the cardboard controls and snake model. The movement of the snake appeared fairly natural because it was heavy enough (850 g) to cling tightly to the forest oor, and the cardboard boxes moved smoothly along the surface of the oor. The cat and dog were in the aviary for similar time periods (between 110 and 182 s); the snake was moved with natural speed (200 s for 16 m). Before the experiments commenced, I tested for any possible reaction of three chicks to the moving loop rope itself without any object attached. No response was obvious. The raptor model and its control box both had a plastic wheel with bearings attached to their top, which ran along a 7.6 m long, plastic-coated clothesline so that both models appeared to ‘ y’ along under the roof, steadily and without any wobbling movements. When a release mechanism in the hide was pulled, the aerial objects emerged from a box 3 m above the ground and ‘ ew’ across the aviary before disappearing into another box, at an angle of 8 degrees. They were exposed for 7-11 seconds, similar to the natural speed of a swooping raptor. Alarm calls and control were both presented in bursts of 4-12 seconds played at random intervals for 2 min, at natural volume and from one loudspeaker at 1 m height. The chicks The brush-turkey was chosen because it is one of the few common megapodes, and eggs can be collected from the wild without harming the population. The behaviour of adult brush- turkeys has been studied intensively in the wild (Jones, 1988a, b, 1990a, b; Birks, 1997, 1999), but to date, chicks have only been observed in captivity (Baltin, 1969; Wong, 1999a, b). Eggs were collected from mounds in Mary Cairncross Park and incubated arti cially (for details see Göth & Jones, in press). After hatching, chicks remained in the incubator until they were dry and were then placed in a thermoinsulated foam box in a dark room, lined with a thick layer of paper towel. They remained in the box for about 42 hours, as this is the mean time they usually spend in the soil before they dig themselves out (Göth, in prep.). To ensure that chicks had only minimum contact with humans, they were released into the aviary by moving one arm through the curtain at the entrance and carefully turning the box upside down. After the trials, the chicks were released into the surrounding rainforest where brush-turkeys occur naturally. Experimental trials Fifty chicks were tested, 10 for each model; each chick was only shown one kind of model. Chicks were always tested when in the aviary alone. The research design consisted of ve consecutive phases: (1) a 15-min BASE LIN E observation of general behaviour, which began when the chick rst moved after it had been placed in the aviary; (2) a T ES T 1 with either the stimulus or control, (3) a 15-min POST- TE ST 1 observation of general behaviour, (4) a T ES T 2 with either the control or stimulus, and (5) a 15-min POST- TE ST 2 observation. Each phase followed immediately after the last. To prevent an order effect, ve chicks receiving one model were shown the control rst and ve were shown the stimulus rst; the order on a given day was chosen randomly. The length of 15 min for one observation period was chosen because this was suf cient to ensure that any response was documented and short enough to avoid habituation or other effects, such as a change in weather. 122 A. GÖTH During the BASE LIN E and both POST- T ES T observations, instantaneous samples of behaviour were taken at 15 s intervals (denoted by an electronic beeper). Recorded behaviours were: (1) Feeding (scratching in the soil or litter, pecking at food, drinking); (2) Resting (standing or sitting motionless with eyes sometimes closed or at least blinking often); (3) Running (fast moving to a new spot; slow walking was also incorporated here because it occurred rarely); (4) Preening (feather preening, scratching, stretching wings or legs, bill wiping, shaking the body, sun- and sandbathing); (5) Crouching (crouching and freezing on the spot, i.e. an immediate change into a motionless posture, were both combined to this category as they were sometimes dif cult to distinguish and intermediate states could occur); (6) Vigilance up (standing with head up, e.g. extended vertically); and (7) Vigilance down (standing or sitting while moving the head and observing the surroundings with the neck drawn in). Data gathering in both T ES TS began as soon as the model entered the aviary or the alarm call/control was rst played. Continuous sampling was used because of the short duration of the TE ST S. Movements and behaviour of the chicks were spoken into a hand-held tape- recorder, with particular reference to ‘line-of-sight’ (between model and chick, not applicable for alarm calls), activity and latency to the rst step (in seconds). ‘Line-of-sight’ was scored as either 1 (chick in view of the predator/control) or 0 (chick hidden as determined by the observer). ‘Activity’ was the behavioural response of the chick, either or both of a posture and an action. An ordinal scale was developed to describe it quickly during the T ES TS , based on previous observations of chicks both in natural and aviary observations. Activity categories were as follows: (0) No visual response: chick continues feeding, resting or preening; (1) Vigilance with head down; (2) Vigilance with head up; (3) Crouching or freezing; (4) Walking away slowly from the model; (5) Running away quickly from the model; (6) Flying up onto a log or tree; and (7) Calling. When more than one category was observed in succession, both their order and duration were recorded. Analyses The data for the TE ST S were recorded as latency to the rst step (in seconds) and proportion of time (in seconds) doing each of feeding, resting, running, preening, crouching, ‘Vigilance head up’ and ‘Vigilance head down’. The duration of the TE ST S varied because the raptor was presented for a shorter period than the ground predators; data were thus converted to percent. Latency to the rst step was assumed to be 100% (of the total time the model was present) if a chick did not move at all and 1% if it ran away as soon as the model appeared. Since it was unknown how the chicks might react to the models, no predictions had been made about an increase or decrease of individual behaviours. Therefore, mixed-factorial ANOVAs were conducted for each behaviour that was observed in the T ES TS , except for those that occurred at very low levels (feeding, resting and preening). In these ANOVAs, ‘treatment’ (stimulus or control) was a within-subject factor because the same chick had been shown a stimulus and its control, whereas ‘model’ ( ve types of stimuli/controls) was a between-subject factor because 10 different birds had been tested for each model. The three null hypotheses tested were that the mean proportion of time that the chicks spent performing each behaviour was not affected by (1) the two levels of the treatment condition; (2) the ve types of models; and (3) any interaction of both. In the POST- TE ST S, both time budget data and latency were measured as proportion of sample points (out of the 60 per observation period) at which each behaviour occurred or PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 123 after which the rst step occurred. For analyses, these were converted to percent. Data for post-predator and post-control behaviour were subtracted from the baseline observation and the resulting two values were used for mixed-factorial ANOVAs in which behaviour was again compared across models and treatments, as described above. Data for latency could not be subtracted because during the baseline observation, the observations had started when the chicks rst moved and no values thus existed; the observed values in the POST- TE ST S were therefore used for mixed-factorial ANOVAs, as described above. All ANOVAs followed arcsine transformation, were two-tailed and based on type III sum of squares. Bonferroni adjustments were used for a post-hoc pairwise multiple comparison to determine which means differed. This test is based on the Student’s t statistic and adjusts the observed signi cance level for the fact that multiple comparisons are made. Non-parametric Wilcoxon signed-ranks tests (two-tailed) were applied for a comparison of behaviour between baseline and POST- T ES T observations as well as for a test for order effects (i.e. order of presentation of models). For the latter tests, data for each behavioural category or latency and for each stimulus and control were grouped by order and compared with Wilcoxon signed-ranks tests. Only 3 of the 32 tests revealed signi cance (p < 0:05), so I concluded that order of presentation was not an important factor, and ignored it in the analysis. All tests were conducted with the Windows95 version of SPSS (6.3), following descrip- tions by Kinnear & Gray (1994); signi cance was accepted at p < 0:05. Results Response in TESTS When presented with any of the ve models during the TESTS, chicks never moved into the two patches of thicket in the aviary and were therefore always in line-of-sight . Some possible behavioural responses were never observed: ying, walking or running towards the model, walking away slowly or calling. Feeding occurred in only three types of tests (during 10% of the snake, 3.3% of the alarm call and 17.1% of the white noise presentation) and resting and preening were represented with no more than 1.8% in any test. All other behavioural traits were observed in all tests and chicks combined them in different ways, most often as ‘Vigilance with head up’ followed by either running or crouching. Figure 1 shows the proportion of these individual behaviours during TESTS with different models. The type of model (cat/control, etc.) had a signi cant effect on the percentage of time spent running (main effect of model factor; mixed factorial ANOVA: F4;45 D 8:40, p < 0:01), crouching (F4;45 D 13:61, p < 0:01) and being vigilant with head up (F4;45 D 3:32, p < 0:05). Only the proportion of ‘Vigiliance with head down’ did not differ between trials (mixed-factorial ANOVA). 124 A. GÖTH Fig. 1. Mean percentage of observation time devoted to behaviour observed during T ES TS with ve different stimuli (black columns) and their controls (white columns). Models and behavioural categories as de ned in the Methods. Bars indicate 95% con dence intervals; N D 10 chicks per model (D stimulus and control). Post-hoc pairwise Bonferroni comparisons revealed that the percentage of running was higher in presence of the dog compared to the raptor and alarm calls (both p < 0:01), and the same applied for the presence of the snake (p < 0:01 compared to raptor and alarm calls). Chicks crouched PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 125 signi cantly more often in the presence of the cat compared to the snake and while the alarm calls were played (both p < 0:05). When the raptor ew overhead, they also crouched signi cantly more than in all other tests (compared to cat p < 0:05; dog p < 0:01; snake and alarm: p < 0:01). The proportion of ‘Vigilance with head up’ was higher during presentation of alarm calls compared to the dog (p < 0:05). There was, however, no obvious difference in response to the four optical stimuli and their controls, as revealed by the lack of a signi cant effect of the treatment condition (main effect of treatment factor; mixed factorial ANOVA: p > 0:05 for crouching and running) and the lack of a signi cant interaction of both (model £ treatment factor; mixed factorial ANOVA: p > 0:05 for crouching and running). Vigiliance with head up, on the contrary, was signi cantly affected by the treatment condition (main effect of treatment factor; mixed factorial ANOVA: F1;45 D 5:56, p < 0:05) and an interaction of model type and treatment condition (treatment £ model factor; mixed factorial ANOVA: F4;45 D 3:34, p < 0:05). The percentage of this behaviour only differed between alarm calls and white noise (Wilcoxon signed-ranks test: z D ¡2:38, N D 10, p < 0:05), not between any other stimuli and their controls. Chicks moved very little or not at all when presented with the raptor and cat compared to the other stimuli (Fig. 2), and accordingly there was a signi cant effect of model type on the latency to the rst step (main effect of model factor; mixed factorial ANOVA: F4;45 D 4:94, p < 0:01; Fig. 2), whereas latency was not affected by the treatment condition (main effect Fig. 2. Latency to the rst step in the TE ST S, expressed as mean percentage of time (of the duration of the test) after which chicks rst moved when presented with different stimuli (black columns) and their controls (white columns). Bars indicate 95% con dence intervals; N D 10 for each model. 126 A. GÖTH of treatment factor; mixed factorial ANOVA: F1;45 D 0:17, p > 0:5) or any interaction of both (model £ treatment factor; mixed factorial ANOVA: F4;45 D 1:32, p > 0:1). Post-hoc Bonferroni tests showed that the mean latency was higher in the presence of the raptor compared to the dog and snake (dog: p < 0:05; snake: p < 0:01). Response in POST-TESTS In the POST- TESTS, after the stimuli had disappeared, most chicks resumed behaviour that they had shown only occasionally during the TESTS, such as resting, preening and feeding (Fig. 3). After the cat had disappeared, chicks crouched more than in the baseline observation (Wilcoxon signed- ranks tests, z D ¡1:99, N D 10, p < 0:05), and after the dog, they ran back and forward more (z D ¡2:67, N D 10, p < 0:01) and fed less (z D ¡2:81, N D 10, p < 0:01). The post-respons e to the raptor model was an increase in crouching (z D ¡2:34, N D 10, p < 0:05) and a decrease in feeding (z D ¡2:04, N D 10, p < 0:05). The behaviour before and after an encounter with the snake or alarm calls did not differ signi cantly (pairwise Wilcoxon signed-ranks comparisons of all behavioural traits in POST- TESTS and BASELINE observations : snake smallest z D ¡1:13, p > 0:1; alarm call smallest z D ¡1:99, p > 0:05). Not only was there a post-response to some stimuli (cat, dog, raptor), but two facts also indicate that behaviour in the POST- TESTS differed between models. First, crouching only occurred after the raptor, cat and the dog or their controls, but never after the snake and alarm calls or their controls; tests showed that this behaviour was affected by model type (main effect of model factor; mixed factorial ANOVA: F4;45 D 2:86, p < 0:01), but that there was again a similar response to the stimulus and its control (mixed factorial ANOVA; main effect of treatment factor: F1;45 D 1:28, p > 0:1; treatment £ model factor: F4;45 D 0:63, p > 0:1). Second, the latency to the rst step was higher after the raptor had disappeared compared to the other stimuli (Fig. 4); tests also revealed that latency was affected by model type (main effect of model factor; mixed factorial ANOVA: crouching: F4;45 D 2:86, p < 0:01; latency: F4;45 D 8:02, p < 0:01), but not by treatment condition (main effect of treatment factor; mixed factorial ANOVA: F4;45 D 0:0004, p > 0:1) or any interaction of both (treatment £ model factor; mixed factorial ANOVA: F4;45 D 0:38, p > 0:1). PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 127 Fig. 3. Mean number of sample intervals in which seven behaviours were observed in trials with ve different models, in the BASE LIN E observation (grey columns), after the presentation of the stimuli (black columns) and their controls (white columns). Bars indicate 95% con dence intervals; N D 10 for each model. Models and behavioural categories as de ned in the Methods; ‘VigUp’ D Vigilance with head up, ‘VigDo’ D Vigilance with head down. One observation period of 15 minutes consisted of 60 sample intervals. Stars give results of pair-wise Wilcoxon signed-rank tests between BASE LIN E (grey) and POST- ST IMU LU S (black) response. 128 A. GÖTH Fig. 4. Latency to the rst step in the POST- T ES TS , expressed as mean number of 15-second sample intervals that passed in one observation period until the chicks rst moved after T ES TS with different stimuli (black columns) and their controls (white columns); one observation period was 15 min and consisted of 60 sample intervals. Bars indicate 95% con dence intervals; N D 10 for each model. Discussion Response to optical stimuli Two-day-old Australian brush-turkey hatchlings, when tested in a large outdoor aviary in the rainforest, respond to optical stimuli, and their response differs with the type of stimulus presented. Other galliformes also crouch or freeze when a ying predator approaches, avoiding detection by relying on their camou age (Dowell, 1986; Evans et al., 1993). Brush-turkeys do the same; their plumage is also very cryptic, a typical feature of most megapode chicks (Jones et al., 1995). They crouch signi cantly more in the presence of a raptor compared to any ground predators, and the strong crouching response is still detectable after the raptor had disappeared. Accordingly, the latency to the rst step is also signi cantly longer during and after the presence of a raptor. Apart from the strong difference in response between aerial and ground predators, brush-turkey chicks also seemed to differentiate between three types of ground predators. When the snake approached, they mostly ran back and forth along the aviary wall, signi cantly more than in the raptor test. As soon as the snake had disappeared, they stopped running. It is unlikely that this response was caused by the fact that the snake was a dummy; experiments on other birds provide good evidence that dummies can elicit appropriate responses to the predators they represent (Curio, 1975; Knight & Temple, 1986). Rather it seems that a snake does not evoke a response that PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 129 is strong enough to be maintained once the stimulus has disappeared. The opposite was the case with the dog and cat. While the cat was in the aviary, the chicks mainly crouched, more than during the snake and alarm test. After the cat had been removed, they remained crouched, signi cantly more than in the baseline observation. The dog mainly evoked more running, in the test itself (more than in the raptor and alarm test) as well as in the post-respons e (compared to baseline). An important nding of this study is that chicks did not respond differently to the four predators and their similar sized and shaped cardboard controls. This indicates that their response was not associated with the species of predator, and that other releasing mechanisms seemed to be involved. For both ground and aerial predators, this might have been size, speed and/or height. The bigger dog and its control might have evoked more running, whereas the smaller size of the cat and control might have been responsible for evoking more crouching. Such a response to size, speed and/or height instead of a certain species of predator makes sense for brush-turkey chicks because, in their evolutionar y history, they co-evolved with a number of predators that differed in size, speed and behaviour, many of which have now become extinct (Flannery, 1994). The lack of a difference in response to stimulus and control is of particular interest for the ying raptor model and its control. It provides further arguments for the discussion about which features of a model are important for evoking a response based on an ‘innate releasing mechanism’. Tinbergen (1963) stated that any bird, or even a cardboard dummy that has a short neck, releases an escape response. In the present experiments, the cardboard box did not have a neck, and the result seems to support the idea that response to an aerial stimulus is primarily based on the relative speed and apparent size of the model (Schleidt, 1961; Evans et al., 1993). In summary, brush-turkey hatchlings seem to be able to differentiate between different sized predators, as obvious in their response, which involves different proportions of crouching and running. Such a ne-scaled discriminatio n among enemies has been shown for a number of adult birds (Kruuk, 1964: Larus argentatus , L. fuscus; Curio, 1975: Ficedula hypoleuca; Grubb, 1977: Fulica americana; Buitron, 1983: Pica pica; Walters, 1990: Vanellus spp.) but, to my knowledge, not for birds soon after hatching. The question has often been approached (e.g. Schaller & Emlen, 1962; Jones, 1980; Mueller & Parker, 1980; Buitron, 1983; Dowell, 1986), but since birds 130 A. GÖTH usually need to be with parents or siblings to behave naturally, it is dif cult to distinguish between responses directly to the predator and responses in uenced by interactions with parents or siblings. In grey partridges, Perdix perdix, and pheasants, Phasianus colchicus, anti-predator behaviour seems to be an automatic response to speci c stimuli, but the birds must learn from their parents how to organise such patterns and for how long to perform them (Dowell, 1986). This was not the case in brush-turkeys , which seem to have an innate response to various predators (or their size). The term innate, though, is heavily discussed in the ethological literature (e.g. Bateson, 1983; McLean & Rhodes, 1991). An ‘innate releasing mechanism’ is de ned as a perceptual mechanism that achieves the identi cation of a stimulus without any prior experience with it (Curio, 1975). Tests on innate releasing mechanisms require a naive animal, that is, one deprived of the very stimuli whose recognition one is going to test. This can be dif cult to arrange, especially when the stimuli are ground and aerial predators. Almost every bird experiences some parts of its environment prior to testing, either immediately after hatching or when still in the egg, where it can hear the parents’ alarm calls. Arti cial incubation could create predator-naive birds, but in almost all bird species, this would also lead to an arti cial situation that could, in return, affect the results. Megapode hatchlings are an exception to these conditions and offer an ideal subject for testing predator-naive birds. Their eggs are buried in the dark and deep underground, thus far away from alarm calls or any other experience with predators. They also have no parents from which they could copy any response after hatching. In the present study, two-day-old chicks were tested because younger ones are not capable of moving properly. In the natural incubation mound, they remain in the soil for about two days and during this time they dry, lose their feather sheaths and nally dig their way up to the surface (Göth, in prep.). It is thus inappropriate to ask at what age in hours brush-turkey hatchlings rst show an innate response to predators, simply because very young chicks are not yet in a physical condition to, for example, run away. They do inch when a human hand approaches them in the incubator (personal observation) but that is all they are capable of. Moreover, it should be noted that there is still no consensus of opinion about the age in hours at which ‘fear behaviour’ rst appears in chicks of the domestic hen (for summary see Rogers, 1995). PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 131 Response to alarm calls Brush-turkey chicks responded appropriately to alarm calls of songbirds that indicated an apparent approaching enemy: all chicks stopped whatever they were doing as soon as the alarm calls were rst played and looked around with their head stretched out vertically (‘Vigilance with head up’), and this response was signi cantly higher than in tests with other stimuli. Addition- ally, alarm calls evoked a signi cantly higher response than their control (white noise), indicating that chicks responded more speci cally to these calls than to optical stimuli. Alarm calls also only evoked a response while they were played and chicks resumed their normal behaviour immediately after. This could, of course, also have been caused by habituation. To avoid such an effect, alarm calls were presented in bursts played at random in in- tervals during the 2 minute long test. Chicks usually looked up when they were played but continued feeding or looking around (‘Vigilance with head down’) soon after the calls had stopped. Also, they quite often stopped look- ing up before the end of the 2 minutes, indicating a slight habituation even while the alarm calls were played. Most young galliform chicks react to alarm or distress calls after hatching, but usually to those of their parents or siblings (Kruijt, 1964; Dowell, 1986; Jones, 1987). Adult megapodes do not utter a speci c alarm call (Jones et al., 1995) and hatchlings almost never call at all (except occasionally when kept in a group or when handled; Göth et al., 1999; pers. obs.). For megapode chicks, alarm calls of other birds have apparently replaced those of the parents and they act as auditory clues in predator-prey interactions. Implications for management This study also aimed at bridging the gap between fundamental and applied research — an aspect largely underrepresented in the behavioural literature (Curio, 1996; Sutherland, 1998; Caro, 1999). Knowledge of the predator recognition ability is important for the application of management plans for endangered megapodes. The results presented here show that megapode hatchlings have the ability to respond to native as well as to introduced predators such as dog and cats, because they react to anything of a certain size or speed. A lack of response should thus not be of concern for the release of hatchlings in captive-release programs. One factor, though, is of crucial importance for such a release: cover. When a ground predator approaches, 132 A. GÖTH the chicks either run away immediately, or they rst crouch and then run. This only makes sense if they are able to reach a safe thicket earlier than the predator reaches them.They do not seem to y when trying to escape and can run fast; in the aviary, they covered distances of 3 m within 2.5 to 4.0 seconds (N D 5 chicks). They are, however, not able to keep up such an energy consuming escape for long (personal observation) and rely, therefore, on cover. Chick mortality in the wild is generally very high (Jones, 1988b; Priddel & Wheeler, 1994; Göth, in prep.), but a recent investigation showed that more chicks survive in areas with large stands of thickets (Göth, in prep.). It is likely that the two other megapodes in Australia, the malleefowl and orange-footed scrubfowl, Megapodius reinwardt, also react to enemies in the same way as the brush-turkey. They, too, co-evolved with similar predators (e.g. cat-like marsupials). Early accounts describe that chicks of the endangered malleefowl ee into cover when large objects move nearby (Frith, 1962). The high percentage of malleefowl chicks killed by foxes (Priddel & Wheeler, 1994, 1996, 1997) may not have been caused by a lack of predator-recognition , but rather by a lack of cover (Priddel & Wheeler, 1999), or the heavy load of the radio-transmitte r attached to some released chicks with harnesses (for a discussion see Göth & Jones, in press). The density of malleefowl is higher in habitats containing a dense and continuous cover, and predation by foxes might be a serious problem only where the vital vegetation cover is destroyed by res, land clearing or grazing (Benshemesh, 1999; Priddel & Wheeler, 1999). A re-introduction program for this endangered species should thus focus on releasing chicks into the right environment, where still available, with particular focus on the availability of a dense ground cover. In other conservation programs, the predator-naive species has to be taught how to recognise predators prior to release (McLean, 1997; McLean et al., 1999; Hölzer et al., 1996), but this does not seem to be necessary for malleefowl chicks. All other endangered megapodes occur on islands (Jones et al., 1995). It can not be predicted whether they do also react to introduced predators, espe- cially ground predators, as hatchlings. Some of these species have co-evolved with native ground predators like monitors, cats and civet-cats, Viverri- dae (Dekker, 1989), but those living on small islands without any native ground predators might not have developed the ability to cope with enemies. A few chicks of one such species, the Polynesian megapode, Megapodius PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 133 pritchardii, however, were kept in aviaries for observations (Göth, 1995). Although their predator-recognitio n ability was not speci cally tested, it was obvious that they were wary and ed from approaching dogs and humans immediately. These results suggest that all megapode hatchlings respond to preda- tors appropriately when released at the age of two days. In conservation projects for endangered megapodes, though, logistics might sometimes force researchers to collect a few chicks before transporting them to the release site. It could be that chicks that are kept in cages for too long before re- lease lose all, or parts of, their ability to respond to predators. Chicks of the hazel grouse, Bonasia bonasia, rock ptarmigan, Lagopus mutus, and rock partridge, Alectoris graeca, needed continual contact with a fox and the alarm calls uttered by their parents to keep up their responses to predators throughout development (Thaler, 1987). Juvenile malleefowl that were re- leased at the age of 3-5 months were all killed, mostly by foxes (Priddel & Wheeler, 1996). To exclude that such results are caused by a lack of predator recognition, chicks should either be released at the age of two days, or eggs should be buried at incubation sites in the area chosen for the chicks, thus enabling them to experience their environment immediately. A release at an earlier age than two days, on the other hand, should also not be considered, because such chicks are not strong enough to move properly. References Baltin, S. (1969). Zur Biologie und Ethologie des Talegalla-Huhnes (Alectura lathami Gray) unter besonderer Berücksichtigung des Verhaltens während der Brutperiode. — Z. Tierpsychol. 26, p. 524-572. Bateson, P.P.G. (1983). Genes, environment and the development of behaviour. — In: Animal behaviour Vol. 3, Genes, development and learning (T.R. Halliday & P.J.B. Slater, eds). Blackwell, Oxford, p. 52-81. Benshemesh, J. (1999). The national Malleefowl recovery plan: a framework for conserving the species across Australia. — In: Proceedings of the Third International Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Ben- shemesh, eds). Zoologische Verhandelingen, Leiden, p. 101-124. Birks, S.M. (1997). Paternity in the Australian brush-turkey Alectura lathami, a megapode bird with uniparental care. — Behav. Ecol. 8, p. 560-568. — — (1999). Unusual timing of copulation in the Australian brush-turkey. — Auk 116, p. 169-177. Brown, C. & Warburton, K. (1997). Predator recognition and anti-predator responses in the rainbow sh Melanotaenia eachamensis. — Behav. Ecol. Sociobiol. 47, p. 61-68. 134 A. GÖTH Buitron, D. (1983). Variability in the response of black-billed magpies to natural predators. — Behaviour 87, p. 209-236. Burger, J. (1998). Antipredator behaviour of hatchling snakes: effects of incubation tempera- tures and simulated predators. — Anim. Behav. 56, p. 547-553. Caro, T. (1999). The behaviour-conservation interface. — Trends Ecol. Evol. 14, p. 366-369. Conover, M.R. & Perito, J.J. (1981). Response of starlings to distress calls and predator models holding conspeci c prey. — Z. Tierpsychol. 57, p. 163-172. Curio, E. (1975). The functional organisation of anti-predator behaviour in the pied y- catcher: a study of avian visual perception. — Anim. Behav. 23, p. 1-115. — — (1993). Proximate and developmental aspects of antipredator behavior. — Adv. Study Behav. 22, p. 135-238. — — (1996). Conservation needs ethology. — Trends Ecol. Evol. 11, p. 260-263. Dekker, R.W.R.J. (1989). Predation and western limits of the megapode distribution. — J. Biogeogr. 6, p. 317-321. — — (1999). The megapode action plan 1995-1999 halfway down the road. — In: Pro- ceedings of the Third International Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Benshemesh, eds). Zoologische Verhandelin- gen, Leiden, p. 151-158. — —, Fuller, R.A. & Baker, G.C. (in press). Megapodes — Status survey and conservation action plan 2000-2004 . — IUCN Switzerland, Gland and UK, Cambridge. — — & McGowan, P.J. (1995). Megapodes: an action plan for their conservation 1995-1999 . — IUCN Switzerland, Gland. Dowell, S. (1986). The development of anti predator responses in gamebird chicks. — Game Conservancy Ann. Rev. 18, p. 93-98. Evans, C.S., Evans, L. & Marler, P. (1993). Effects of apparent size and speed on the response of chickens, Gallus gallus, to computer-generated simulations of aerial predators. — Anim. Behav. 46, p. 1-11. Flannery, T.F. (1994). The future eaters. — Reed, Melbourne. Frith, H.J. (1962). The Mallee fowl. — Angus & Robertson, Sydney. Gallup, G.G.Jr., Cummings, W.H. & Nash, R.F. (1972). The experimenter as an independent variable in studies of animal hypnosis in chickens (Gallus gallus). — Anim. Behav. 20, p. 166-169. Göth, A. (1995). Some aspects of the ontogeny of the Polynesian Megapode (Megapodius pritchardii, Megapodiidae) (Summary of diploma-thesis). — Megapode Newsl. 9, p. 4- 6. — — & Jones, D.N. (in press). Transmitter attachment and its effect on superprecocial Australian brush-turkey hatchlings. — Wildl. Res. — —, Vogel, U. & Curio, E. (1999). The acoustic communication of the Polynesian megapode Megapodius pritchardii G.R. Gray. — In: Proceedings of the Third Inter- national Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Benshemesh, eds). Zoologische Verhandelingen, Leiden, p. 37-52. Grubb, T.C.Jr. (1977). Discrimination of aerial predators by American coots in nature. — Anim. Behav. 25, p. 1065-1066. Hartzler, J.E. (1974). Predation and the daily timing of sage grouse leks. — Auk 91, p. 532- 536. Hölzer, C., Bergmann H.-H. & McLean, I.G. (1996). Training captive-raised, naive birds to recognise their predator. — In: Research and captive propagation (U. Ganslosser, J.K. Hodges & W. Kaumanns, eds). Filander Verlag, Fürth, p. 198-206. PREDATOR-RECOGNITION IN MEGAPODE HATCHLINGS 135 Jones, D.N. (1988a). Construction and maintenance of the incubation mounds of the Aus- tralian brush-turkey Alectura lathami. — Emu 88, p. 201-218. — — (1988b). Hatching success of the Australian brush-turkey Alectura lathami in South- East Queensland. — Emu 88, p. 260-263. — — (1990a). Male mating tactics in a promiscuous megapode: patterns of incubation mound ownership. — Behav. Ecol. 1, p. 107-115. — — (1990b). Social organization and sexual interactions in Australian brush-turkeys Alectura lathami: implications of promiscuity in a mound-building megapode. — Ethology 84, p. 89-104. — — (1999). What we don’t know about megapodes. — In: Proceedings of the Third In- ternational Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Benshemesh, eds). Zoologische Verhandelingen, Leiden, p. 159-168. — —, Dekker, R.W.R.J. & Roselaar, C.S. (1995). The megapodes. — Oxford University Press, Oxford. Jones, R.B. (1980). Reactions of male domestic chicks to two-dimensional eye like shapes. — Anim. Behav. 28, p. 212-218. — — (1987). Social and environmental aspects of fear in the domestic fowl. — In: Cognitive aspects of social behaviour in the domestic fowl (R. Zayan & I.J. Duncan, eds). Elsevier, Amsterdam, p. 113-149. Kinnear, P.R. & Gray, C.D. (1994). SPSS for Windows made simple. — Lawrence Erlbaum Associates, East Sussex. Knight, R.C. & Temple, S.A. (1986). Methodological problems in studies of avian nest defence. — Anim. Behav. 34, p. 561-566. Krätzig, H. (1939). Untersuchungen zur Biologie und Ethologie des Haselhuhnes während der Jugendentwicklung. — Ber. Ver. Schles. Ornithol. 24, p. 1-25. Kruijt, J.P. (1964). Ontogeny of behaviour in Burmese red jungle fowl (Gallus gallus spadiceus) Bonnaterre. Behaviour, Suppl. XII, Leiden, Netherlands. Kruuk, H. (1964). Predators and anti-predator behaviour of the black-headed gull. — Behaviour 11, p. 1-129. McLean, I.G. (1997). Conservation and the development of behavior. — In: Behavioral approaches to conservation in the wild (J.R. Clemmons & R. Buccholz, eds). Cambridge Univ. Press, Cambridge, p. 132-156. — — & Rhodes, G. (1991). Enemy recognition and response in birds. — Curr. Ornith. 8, p. 173-211. — —, Hölzer, C. & Studholme, B.J. (1999). Teaching predator-recognition to a naive bird: implications for management. — Biol. Cons. 87, p. 123-130. Mueller, H.C. & Parker, P.G. (1980). Naive ducklings show different cardiac response to hawk than to goose models. — Behaviour 74, p. 101-113. Murphy, L.B. (1978). The practical problems of recognising and measuring fear and explo- ration behaviour in the domestic fowl. — Anim. Behav. 26, p. 422-431. Priddel, D. & Wheeler, R. (1994). Mortality of captive-raised Malleefowl, Leipoa ocellata, released into a mallee remnant within the wheat-belt of New South Wales. — Wildl. Res. 21, p. 543-552. — — & — — (1996). Effect of age at release on the susceptibility of captive-reared Malleefowl Leipoa ocellata to predation by the introduced Fox Vulpes vulpes. — Emu 96, p. 32-41. 136 A. GÖTH — — & — — (1997). Ef cacy of fox control in reducing the mortality of released captive- reared Malleefowl Leipoa ocellata. — Wildl. Res. 23, p. 469-482. — — & — — (1999). Malleefowl conservation in New South Wales: a review. — In: Proceedings of the Third International Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Benshemesh, eds). Zoologische Verhandelingen, Leiden, p. 125-141. Rogers, L. (1995). The development of brain and behaviour in the chicken. — CAB International, Wollingford. Schaller, G.B. & Emlen, J.T. (1962). The ontogeny of avoidance behaviour in some precocial birds. — Anim. Behav. 10, p. 370-381. Schleidt, W.M. (1961). Reaktionen von Truthühnern auf iegende Raubvögel und Versuche zur Analyse ihrer AAM’s. — Z. Tierpsych. 18, p. 534-560. Sutherland, W.J. (1998). The importance of behavioural studies in conservation biology. — Anim. Behav. 56, p. 801-809. Thaler, E. (1987). Studies on the behaviour of some phasianidae-chicks at the Alpenzoo- Innsbruck. — J. Science Fac Chiang Mai Univ. 14, p. 135-149. Tinbergen, N. (1963). On aims and methods of ethology. — Z. Tierpsychol. 20, p. 410-433. Walters, J.R. (1990). Anti-predator behavior of lapwings: eld evidence of discriminative abilities. — Wilson Bull. 102, p. 49-70. Wong, S. (1999a). Visual predator recognition in Australian brush-turkey (Alectura lathami) hatchlings. — In: Proceedings of the Third International Megapode Symposium, Nhill, Australia, December 1997 (R.W.R.J. Dekker, D.N. Jones & J. Benshemesh, eds). Zoologische Verhandelingen, Leiden, p. 61-74. — — (1999b). Development and behaviour of hatchlings of the Australian brush-turkey Alectura lathami. — Unpubl. PhD thesis, Grif th University, Brisbane, Australia.