1. Introduction
The serval (
Leptailurus serval) is a small felid that is native to various regions of sub-Saharan Africa [
1,
2,
3,
4]. The species is a wetland specialist that navigates between marshes and reed beds in search of its main food source, which is small rodents [
1,
2,
3,
4]. A characteristic hunting technique where the serval relies on its superior sense of hearing to locate prey in tall grasses, after which it then leaps and pounces on the prey item, makes it a highly successful predator [
2,
4]. Like most other felids, the serval is a solitary species [
4]. The female commonly gives birth to a litter of 1–3 kittens in summer and cares for her young for a considerable amount of time after they become mobile [
4].
The serval is currently classified as “least concern” [
5], although wild populations have recently declined in parts of their range due to secondary poisoning by consumption of poisoned rodent prey [
3]. Wild servals are also subjected to attacks by domestic dogs or are shot by local farmers to prevent predation on domestic poultry [
6]. Fortunately, the species is well represented in captivity [
7] and attempts at re-introducing captive individuals into the wild have proven successful [
2]. The reproductive success rate among captive servals is, however, relatively low [
8]. As with many small felids, very little is known about the welfare of servals in captivity, since most research efforts to date have been dedicated to the larger and more conspicuous felids,
Panthera sp. [
9,
10]. Welfare-oriented research in these species has mainly focused on implementing and assessing the effects of various enrichment strategies [
11,
12,
13,
14], exhibit designs [
15,
16], and housing constellations [
17,
18]. Although this research has often led to measurable improvements in welfare, similar studies in small cats are comparatively scarce (however, see References [
19,
20,
21]).
Additionally, only a handful of studies to date have addressed the topic of visitor–animal interaction and its welfare consequences for captive felids, even though these animals may be exposed to unfamiliar humans on a daily basis [
22], and close visitor interaction with felids is becoming increasingly common in zoos worldwide [
23]. As such, a range of interactive programs are now offered with various felids. A common feature is to grant visitors access to off-limit areas where they interact with and tong-feed a big cat, most commonly a lion or a tiger, through a protective barrier [
24]. Encounters with small felids, including cheetahs,
Axinonyx jubatus, and servals, are often more tactile in nature and commonly allow visitors to pat and have their photos taken with a cat, or engage in an interactive walk together with a cat and its keeper [
24]. In addition, servals are commonly featured as animal ambassadors in serendipitous encounters or educational workshops in zoos [
24].
Although encounter programs with captive felids are now commonplace in many zoos, Szokalski et al. [
23] consider this a controversial practice, given the solitary and elusive nature of many felids. There is, however, not enough empirical data available to suggest that close visitor interaction may exert a negative welfare impact on participating animals. Szokalski et al. [
23] studied the effects of interactive programs on the behaviour of captive lions,
Panthera leo, and cheetahs at two Australian zoos and observed high levels of stereotypic pacing among lions prior to encounters. The encounters involved protected contact feeding, and it was suggested that food anticipation, rather than stress caused by close visitor interaction, may have been responsible for this effect. Cheetahs, on the other hand, frequently expressed signs of affiliative behaviour towards both visitors and keepers during interactive walks, suggesting that it may have been a positive welfare experience for these individuals [
23]. Another Australian research team studied fluctuations in faecal glucocorticoid (FGM) concentration in tigers,
Panthera tigris, participating in walks, interactive presentations, and guest photo opportunities at two different zoos, and found that program animals had higher overall concentration of FGMs compared to non-participating animals at one institution, but the opposite trend was observed at the second zoo [
25]. The authors suggest that variation in the level of conditioning and familiarity with the public may have been the reason for this disparity. There were, however, no behavioural observations that could further strengthen this claim, since the main focus of this study was to validate a physiological assay [
25].
Given their widespread occurrence in zoos and their popular role as program animals, studies investigating the welfare impacts of encounters with captive servals is clearly worthwhile, in order to support the continued involvement of this species in interactive programs. Such research would optimise the care and welfare of individual animals and contribute to our understanding of visitor–animal interaction in small exotic felids. The aim of the current study was to explore the overall impact of encounters on the behaviour, physiology, and potentially short-term welfare of individual zoo-housed servals. Specifically, the aim was to determine whether potential welfare impacts were affected by:
It was hypothesized that variation in encounter frequency, as well as type of encounter, would elicit changes in serval behaviour and physiology that could be indicative of a welfare impact, though it could not be predicted whether this impact would be positive or negative due to the paucity of information on this topic. To our knowledge, our study was the first of its kind to investigate a potential cause–effect relationship between visitor interaction and behavioural and physiological welfare measures in captive servals.
2. Materials and Methods
All animal procedures in the current study were approved by the Zoos Victoria Research and Animal Ethics Committee (ZV16003).
2.1. Study Animals—Housing and Husbandry Routine
Study animals included two adult servals housed at Werribee Open Range Zoo (WORZ), which is situated 35 km southwest of Melbourne, Victoria, Australia. The servals, named Nanki and Morilli, were both captive-bred females from the same litter. They were born at Mogo Zoo, New South Wales, Australia, in December 2008, and were relocated to WORZ in March 2009. Although parent-reared from birth, once they arrived at WORZ, they were bottle-fed by keepers while being introduced to solid foods and were subsequently weaned at around six months of age. Their current feeding regime included two meals per day: one morning and afternoon feed consisting of 2–3 mice or day-old chicks each, and a rabbit or chicken leg each. In addition, each serval received two small portions of diced red meat per day, which was usually given as a reward during training sessions or visitor interaction.
The servals were housed off public display and could only be seen by the visiting public during presentations or behind-the-scenes (BTS) encounters. The servals were housed solitarily and alternated between an open yard (
Figure 1b) and three adjacent pens (
Figure 1a). The keepers swapped the cats’ housing on a daily basis. The open yard had a total area of approximately 75 m
2 and a ground cover of sand and mulch. The yard was interspersed with logs, elevated platforms, and climbing structures, had a covered nest with a straw bed, a drinking trough, and a designated area for visitor interaction (
Figure 1b). The pens had a total combined area of approximately 36 m
2 and a ground cover of synthetic turf mats and sand. The pens also had elevated shelves, burlap hammocks, cat tunnels, drinking troughs, and heated beds for overnight rest (
Figure 1a).
2.2. Visitor Interaction Program
Shortly after the servals arrived at WORZ as young kittens, the keepers began conditioning the cats for becoming ambassador animals in an interactive program. The animals were taught to follow basic instructions such as recall and sitting on command and were given food rewards and verbal praise to reinforce such behaviours. At the time of the study, the servals had participated in the program for seven years, and were able to perform a complex series of behaviours on command, with the aim of highlighting some of the serval’s natural foraging techniques to visitors (i.e., beam walks to simulate capture of roosting birds in a tree, leaps and pounces to simulate capture of small rodents, and retrieving a meat reward out of a pond to simulate capture of fish and frogs).
The current program involves a daily presentation and a BTS encounter that takes place four days a week. The servals alternated in participating in the daily presentation, but both individuals usually participated in BTS, although one at a time, so when one cat interacted with visitors, the other was held in the off-limit area until the keeper swapped them over. The daily presentation was included in the general zoo admission and took place at 11:00 am in a designated presentation space adjacent to the serval enclosures. The presentation typically attracted large numbers of visitors (maximum 250 people), but the audience did not interact with the serval (i.e., no feeding, touching) and the front row was seated approximately 2 m away from the front of the stage. The participating serval was escorted on a leash by its keeper to the presentation space. Once secured inside the space, the serval was encouraged to undertake a routine training session in front of the audience. The serval was rewarded with red meat for its efforts. The presentation typically lasted for 10–15 min and involved an educational message from the keeper about serval biology, captive management, and conservation of African wildlife.
The behind-the-scenes encounter incurred an additional fee for visitors and took place at 1:30 pm every Tuesday, Thursday, Saturday, and Sunday, in a designated interaction space inside the serval yard (
Figure 1b). A small group of visitors (maximum 6 people) were escorted by a zoo volunteer into the yard where they got to meet the keeper and the servals. Children had to be at least 8 years old to participate, and participants were to stay seated and follow the keeper’s instructions at all times. The encounter involved a training session that was typically longer and involved more spectacular leaps than the presentation. The visitors were also given the opportunity to interact with the serval up close, by having their photograph taken with a cat, gently stroke the cat on its back and chest if the cat allowed, and have the cat lick a small quantity of cream cheese from the visitor’s finger. The encounter typically lasted for 30–40 min and involved a similar educational message as seen in the presentation.
On days when servals did not participate in any visitor interaction, they typically received a one-on-one training session with their keeper inside the serval yard or were taken for a walk by their keepers in the off-limit areas.
2.3. Experimental Design
Using a repeated measures design, the present study imposed changes to the regular program by implementing weekly blocks of four different treatments. The treatments were designed with the aim of separating the effects of presentations and BTS, as visitor number and visitor–animal proximity was markedly different between the two. Hence, the treatments involved participation in either presentations or BTS, the two combined, or no involvement in any visitor interaction (
Table 1). Treatments were imposed for seven consecutive days and changed over every Friday of the week. Each study animal was subjected to each treatment for a total of three weeks, resulting in a total study period of 12 weeks. The two cats alternated between the four treatments, so that each animal was subjected to a different treatment each week. Treatments 1 and 2 always occurred together, i.e., when one cat undertook presentations, the other cat undertook BTS (
Table 1). Likewise, treatments 3 and 4 always occurred together, so that one cat undertook both presentations and BTS while the other cat did not participate in any visitor interaction (
Table 1). The reason for this was to cause minimal interference with the regular program, by ensuring that at least one cat would be available for either presentations or BTS in any given week. The order of the treatments was randomised prior to the onset of the study. Apart from changes to the level of visitor interaction, no other alterations to husbandry and housing were implemented during the study period. Keepers were encouraged to continue taking the cats for walks and undertake one-on-one training sessions when deemed necessary, so that the amount of training and physical activity remained relatively constant across treatments. The study was undertaken between May and September 2016.
2.4. Behavioural Observations
Qualitative behavioural assessments, where behaviours that could be indicative of both positive and negative welfare are monitored on an individual basis, is a widely used approach in welfare-related research [
26,
27,
28,
29]. In the present study, behaviour was monitored over the last three treatment days of each study week (Tuesday–Thursday) using cameras installed in the serval enclosures. A total of 14 cameras (PACOM Dome Close Circuit Television (CCTV)) were used, and they were positioned accordingly: 2–3 cameras in each serval pen, and seven cameras in the yard. The cameras were all connected to a DVR (PACOM Digital Video Recorder, Pacific Communications, Melbourne, Australia) and an external viewing source positioned in a crate in the nearby airlock. The software I-Watch (version 1.2.0, Shield Technology) was used for all subsequent video analysis. The cameras were set to record throughout each observation day. Four designated recording sessions were then sub-sampled from the footage, with the aim of capturing behaviour immediately prior to, during, and after visitor interaction, as well as in the morning and afternoon (
Table 2). During presentations and BTS, only the behaviour of the non-participating cat was monitored (
Table 2), since the participating cat was either in the presentation space or participating in BTS at this time. Since BTS was of longer duration than the presentation (typically lasted for 30–40 min), behaviour of the non-participating cat was monitored for the first 15 min of the encounter only. For the purpose of analysis, the four sessions were collapsed into two daily observation blocks (
Table 2).
Behaviour was monitored by instantaneous scan sampling every 60 s [
30]. Behaviour was not recorded if the keeper was in the enclosure at the time of scanning, as their presence tended to affect the cats’ behaviour. A comprehensive ethogram based on previously published data [
27] was used as a basis for scoring serval behaviour. The final ethogram used in this study (
Table 3) contained four different behavioural categories: passive, active, maintenance, and abnormal repetitive behaviours. At each observation session, the total number of behaviours that the cat was engaged in was referred to as behavioural diversity.
2.5. Adrenocortical Activity
Short-term as well as prolonged exposure to a stressor is often reflected as a change in circulating glucocorticoids, hence, cortisol and its associated metabolites is usually the preferred hormone of choice when assessing physiological stress responses [
27,
32]. In the current study, adrenocortical activity was monitored throughout the study period by analysing variation in excreted levels of faecal glucocorticoid metabolites (FGMs). This technique has been validated previously in a variety of mammal species, including felids [
33,
34], and is considered a reliable and non-invasive approach to measuring patterns of adrenocortical activity. Individually identifiable faecal samples were collected during each treatment day, and 1–2 days post treatment, to account for excretion lag time [
35,
36]. Typically, the cats defecated once a day, though occasionally somewhat more or less frequently. Each study day, keepers were instructed to collect all faeces from both individuals. As soon as the keeper in charge became aware that a cat had defecated, the keeper collected the entire scat, and placed it in a plastic zip-lock bag labelled with animal ID, date, and collection time. Samples could be anywhere between <1 h–15 h old upon collection (the latter applied if the cat(s) had defecated overnight and the sample was not collected until the next morning). Immediately upon collection, samples were transferred to a freezer (−20 °C) where they were stored until extraction and analysis. FGMs were extracted using the ethanol-vortex method [
33,
34] by adding 4 mL of 80% ethanol to 0.5 (±0.01) g of homogenised, wet faeces placed in 5 mL polypropylene vials. Capped vials were vortexed and placed on an orbital shaker overnight. The following day, samples were centrifuged for 5 min at 5000 rpm. The supernatant was then decanted into 1 mL microcentrifuge vials and assayed immediately. The total number of samples analysed per treatment was as follows: Treatment 1 (Presentations only)—23 samples, Treatment 2 (BTS only)—22 samples, Treatment 3 (Presentations + BTS)—27 samples, Treatment 4 (No interaction)—23 samples. Based on these samples, a mean value was generated for each treatment, in order to compare potential changes in FGM across treatments.
FGMs were measured using a group-specific glucocorticoid enzyme immunoassay that had previously been validated for felids in general [
36,
37] and servals in particular [
38]. The corticosterone antibody and corresponding horseradish peroxidase conjugate were both obtained from J. Brown (Smithsonian Institution, Washington, DC, USA; Lab code Cs6). Assay protocols followed previously described methods [
36]. Briefly, 96-well microtitre plates were coated with 150 µL of goat anti-rabbit IgG (2 μg/mL). Immediately prior to use, plates were washed three times. 50 µL of standard, control, or diluted faecal extract were added to each well, immediately followed by 50 µL of horse-radish peroxidase-conjugate (1:80,000) and antibody (1:100,000). The plate was then incubated for two hours at room temperature while shaking, then washed four times to remove unbound steroids. Subsequently, 150 µL of substrate solution was added to each well, and the plate was then incubated at room temperature for approximately 45 min, until the optical density of the maximum binding wells was >0.7. Optical density was read immediately on an Anthos 2010 plate reader (Anthos Labtec Instruments, Austria) at a wavelength of 450 nm. All samples were assayed in duplicate. FGM concentrations (expressed as ng/g wet faecal weight) were calculated from the standard curve using Skanlt RE 4.1 software. To monitor precision and reproducibility, low (∼70% binding) and high (∼30% binding) control samples were run on each plate (four plates were run in total). Inter-assay coefficients of variation (CVs) for low and high controls were 2.6% and 4.5%, respectively. Intra-assay CV was <15%. The assay was biochemically validated by demonstrating parallelism between a serially diluted sample pool and the standard curve.
2.6. Statistical Analysis
The data obtained in this study was analysed using GenStat version 16. To determine the effect of treatment on behaviour (
Table 4) and adrenocortical activity (
Section 2.5), a general analysis of variance was used, where individual animal, treatment (1—Presentation, 2—Behind the scenes, 3—Presentations + BTS, 4—No interaction) and session (observation block 1 or 2;
Table 3) were the treatment structure (
Table 1), and mean number of scans spent in passive, active, maintenance, and abnormal repetitive behaviours as well as concentrations of faecal glucocorticoid metabolites were the Y-variates. Additionally, when examining the effect of housing on behaviour, treatment and housing (pens or yard) were used as independent variables, and the above-mentioned behavioural measures were used as dependent variables in a general analysis of variance. In the analysis, degrees of freedom (d.f.) were: Treatment—3, Individual—1, Session—1, and Total d.f.—15.
In all analyses, a post-hoc Fisher’s least significant difference (LSD) was conducted to determine significance of differences between means for the independent variables. A significance level of 0.05 was used in all analyses.