Non-lethal control of wildlife: using chemical repellents as feeding deterrents for the European badger Meles meles



    Corresponding author
    1. Wildlife Conservation Research Unit, Oxford University, Tubney House, Abingdon Road, Tubney, Abingdon OX13 5QL, UK
      Sandra E. Baker, Wildlife Conservation Research Unit, Oxford University, Tubney House, Abingdon Road, Tubney, Abingdon OX13 5QL, UK (fax +44 1865393101; e-mail
    Search for more papers by this author

    1. Wildlife Conservation Research Unit, Oxford University, Tubney House, Abingdon Road, Tubney, Abingdon OX13 5QL, UK
    Search for more papers by this author

    1. Environmental Risk Assessment Team, Central Science Laboratory, Sand Hutton, York YO41 1LZ, UK
    Search for more papers by this author

    1. Wildlife Conservation Research Unit, Oxford University, Tubney House, Abingdon Road, Tubney, Abingdon OX13 5QL, UK
    Search for more papers by this author

Sandra E. Baker, Wildlife Conservation Research Unit, Oxford University, Tubney House, Abingdon Road, Tubney, Abingdon OX13 5QL, UK (fax +44 1865393101; e-mail


  • 1Non-lethal methods of controlling wildlife foraging damage may offer conservation, ethical, legal and efficacy advantages over lethal control. Chemical repellents present a potential non-lethal approach, but have not been adequately researched in natural environments. Many previous studies have been poorly designed and a lack of data on individual behavioural responses has limited the practical development of repellents. We aimed to identify effective repellents for resolving feeding conflict with wild mammals, using European badgers Meles meles as models.
  • 2We tested the relative efficacy of capsaicin, cinnamamide and ziram, in a multichoice paradigm, using remote video-surveillance to obtain detailed behavioural observations of known free-ranging individuals. Treatment nights were alternated with control nights over 56 nights.
  • 3Badgers discriminated precisely between the four treatments, demonstrating a clear preference for untreated baits, followed by cinnamamide and capsaicin (in no particular order) and then ziram.
  • 4All untreated baits, and baits treated with capsaicin or cinnamamide, were eaten throughout the trial.
  • 5Ziram baits were fully consumed on treatment nights 1 and 2. Ziram consumption then declined to zero between treatment nights 3 and 9, this coinciding with a sharp rise in bait patch rejection. This ‘learning curve’ peaked at treatment night 7. We conclude that badgers developed conditioned taste aversion towards ziram-treated baits at this point. Ziram bait consumption was practically zero over the last 20 treatment nights (40 trial nights) and individuals avoided ziram baits, without sampling, for the last 12–22 treatment nights (24–44 trial nights). Observed changes in badger behaviour suggested that avoidance at a distance was facilitated by odour cues.
  • 6Synthesis and applications. This study provides proof of the concept that ziram has clear potential for reducing badger feeding damage through conditioned taste aversion to an odour. Our detailed observations allowed us to elucidate the behavioural mechanism involved, crucial for directing future development of this approach, thus demonstrating the importance of studying individual responses in wildlife management research. Second-order conditioning, such as this, might be applicable to managing other wild mammals. The next step will be to develop a strategy for use in wildlife damage situations.


Non-lethal methods of controlling wildlife foraging damage may offer conservation, ethical and legal advantages over lethal control (Liss 1997; Ormerod 2002; Macdonald & Baker 2004). A non-lethal approach may also prove more efficient than lethal control, by avoiding the density-dependent consequences that can follow culling (Baker & Macdonald 1999; Shi et al. 2002). However, non-lethal control methods have not been adequately researched in natural environments (Liss 1997; Barlow 2000). Chemical repellents could offer a benign alternative to culling (Baker & Macdonald 1999; Cowan, Reynolds & Gill 2000). Although there have been a number of studies on the use of repellents for wildlife management (Mason 1997), many studies with wild animals have been poorly designed (Reynolds 1999) or have relied on circumstantial evidence of the target species’ response, for example bait consumption, animal signs or sightings in the vicinity (Gustavson et al. 1976; Hanners & Southern 1979; Nicolaus et al. 1989c; Conover 1990). An understanding of individual behaviour is imperative to our ability to manage populations effectively (Alonzo, Switzer & Mange 2003). Indeed, a lack of information on individual behavioural responses has limited the practical development of repellents for managing wildlife (Rogers 1974; Gustavson et al. 1976).

Animals have evolved a range of behavioural mechanisms for avoiding toxins (Garcia & Hankins 1975, 1977) and chemical feeding repellents exploit these mechanisms for pest management (Gill et al. 1999). In this study, our use of the word ‘repellent’ is generic and should not be construed to mean that all substances that ultimately produce avoidance of referent food do so by the same means. Indeed, noxious substances evoke an innate aversion to an odour or taste, whereas conditioned taste aversion (CTA) agents create a CTA as a result of associating an adverse post-ingestional effect (malaise or emesis) with the taste (and in some cases odour) of the food. Some repellents do, however, create both innate and conditioned aversions (Conover 1984; Watkins et al. 1994; Gill et al. 1999; Macdonald & Baker 2004).

European badgers Meles meles L. make good models for studies on conflict resolution. Badgers are found across northern Europe, European Russia and south China, and, in Britain, they epitomize a set of circumstances driving research into non-lethal control. British farmers commonly cite badgers as an agricultural pest (Moore et al. 1999; Baker & Macdonald 2000); badgers are opportunistic omnivores (Moore et al. 1999), numerous in England and Wales (Cresswell et al. 1989) where they cause an estimated £6·5–12·5 million of direct damage to crops annually (Moore et al. 1999). They are also implicated in the spread of bovine tuberculosis (Tb) to cattle (Krebs 1997). However, they are heavily protected by legislation (Protection of Badgers Act 1992) instigated primarily to prevent badger-baiting (Harris et al. 1994) and, despite being unpopular with sections of the rural community, they remain extremely popular with the public.

Badger-damage surveys have concluded that future work should target non-lethal methods of limiting crop damage by badgers (Wilson 1993; Moore et al. 1999). In 1996, interrogation of the UK government's Agricultural Development and Advisory Service's COSTER (Computerized Summary of Technical Reports) database revealed that the majority of badger-damage complaints might be solved using repellents. The only repellent approved for use against badgers, in the UK, was bone oil (30%w/w bone oil, Renardine 72–2™; Roebuck Eyot Ltd, Bishop Auckland, Co. Durham, UK) [Pesticide Safety Directorate (PSD), personal communication]. Renardine™ has a noxious odour and is intended to act as an area repellent. However, it has proved ineffective in reducing badger-feeding damage (Macdonald, Atkinson & Brown 1990) and, following the UK Pesticide Review, Renardine 72–2™ was suspended (and hence withdrawn from sale) on 24 March 2003 (PSD, personal communication). In trials, seven other putative odour repellents also failed to deter feeding in badgers (S. E. Baker, unpublished data).

Identification of a repellent's mode of action is important on the grounds of cost, environmental safety and welfare (Clark, Bryant & Mezine 2000). Furthermore, it is recognized that understanding individual behaviour is imperative in devising appropriate means of conflict resolution (Alonzo, Switzer & Mange 2003). Nevertheless, most repellency studies with wild animals have relied on circumstantial, rather than behavioural, evidence. A few notable exceptions have demonstrated the utility of studying detailed behavioural responses among individually known, free-ranging animals in determining the effect of a repellent (e.g. racoons Procyon lotor, Nicolaus, Hoffman & Gustavson 1982; Semel & Nicolaus 1992; red-winged blackbirds Agelaius phoenicus, Nicolaus & Lee 1999; black bears Ursus americanus, Ternent & Garshelis 1999). Our work adds a further level of detail to the study of behaviour aversion (e.g. every rejection event) through use of an autonomous, infra-red, remote, video-surveillance system, which facilitates reliable, continuous, close-up (4 m) filming without incurring human observer effects (Stewart, Ellwood & Macdonald 1997). Our goal was not only to test repellents as potential feeding deterrents, but also to reveal something of the mechanisms involved. We conducted a multiple-choice ‘cafeteria-style’ experiment (often used with captive animals; Andelt, Burnham & Manning 1991; Andelt, Burnham & Baker 1994) to assess the relative efficacy of three food-based repellents with free-ranging, individually marked, wild badgers. This is the first known test of food-based repellents with badgers.

We tested three compounds, selected from those used in other studies: cinnamamide, capsaicin and ziram. Cinnamamide is a synthetic derivative of plant defence compounds (Watkins 1996), an irritant and broad-spectrum repellent for herbivores, creating both innate and conditioned responses (Watkins, Gurney & Cowan 1998; LD50 orally in rats 1·4 g kg−1; Merck 1989). Capsaicin is the pungent principal component of Capsicum peppers, an irritant, thought universally repellent to mammals (Mason et al. 1991), being innately aversive (Andelt, Burnham & Baker 1994) and capable of causing post-ingestional effects on feeding behaviour (Ritter & Taylor 1989; LD50 orally in mice 1·6 g kg−1, Merck 1989). ‘Hot sauces’ containing capsaicin are used in the USA as repellents against mammals (Andelt, Burnham & Baker 1994). Although the active substance, capsaicin, is not registered as a repellent in the UK, it is registered for use in foods, and several products containing pepper are approved as repellents for amateur use (PSD, personal communication). Ziram (zinc dimethyl dithio-carbamate) is a fungicide registered in the UK as a repellent against birds, deer, hares Lepus europaeus L. and rabbits Oryctolagus cuniculus L. Ziram has known aversive activity against mammals, has irritant properties (Merck 1989) and, like other carbamate fungicides, causes CTA (Mason & Clark 1992; LD50 orally in rats 1·4 g kg−1, Merck 1989).

Materials and methods

The experiment took place on rough pastureland within Wytham Woods, Oxfordshire, UK (1°20′W 51°46′N; UK National Grid SP 46 08), from 19 July to 19 November 1996, at the convergence of three separate badger social group home ranges. This maximized the number of animals attending the feeding site. Badgers were routinely monitored as part of a parallel population study (Macdonald & Newman 2002). Badgers were trapped at focal setts, weighed, sexed and individually marked using either Nyanzol-D (matt black animal dye; Albanil Dyestuffs, Jersey City, NJ, USA) or fur-clipping (Stewart & Macdonald 1997) according to their moult condition. Both types of marking are conspicuous under infra-red light.


Beta Puppy 1–6 months™ (since renamed Beta Bioplus Puppy; Nestlé UK Ltd, Croydon, UK) pelleted food was used as the substrate for coating. These pellets are readily accepted by badgers, are reasonably uniform in size and composition, and have sufficient surface area to allow application of an adequate coating of chemical. Importantly, pelleted bait encouraged animals to feed in situ, rather than carry the food out of camera-shot. This allowed us to observe all feeding activity.

We tested repellent concentrations based on previous work on birds (ziram), wood mice Apodemus sylvaticus and house mice Mus musculus (cinnamamide) and grey squirrels Sciurus carolinensis (capsaicin) (Cummings et al. 1994; Gurney et al. 1996; Central Science Laboratory 1998). Formulations of ziram (1·5% a.i. w/w pellet; AAProtect™ formulation, Dalgety Agriculture Ltd, Reading, UK), cinnamamide (1% a.i. w/w pellet in methanol; Sigma-Aldrich Chemical Co., Gillingham, UK) and capsaicin (0·05% a.i. w/w pellet in diethyl ether; Sigma-Aldrich Chemical Co.) were applied topically to (the surface of) food pellets using an industrial food mixer. This process produced an evenly coated batch, with little or no damage. Treated pellets were air dried and stored in sealed containers until required. Each carrier was appropriate to the repellent and formed part of the repellent treatment, i.e. we tested ziram in the form of AAprotect™, cinnamamide with methanol, and capsaicin with diethyl ether. Control baits were untreated.

experimental design

Observational data were recorded under infra-red illumination (wavelength 880 nm) at an intensity outside the mammalian visual range (Lythgoe 1979). Infra-red sensitive video-surveillance equipment (Stewart, Ellwood & Macdonald 1997) effectively allowed animals to participate in total darkness. The field of view (6·5 × 8 m) contained six 45 × 45-cm concrete paving slabs, arranged in a hexagon (centres of adjacent feeders 1·77 m apart) and serving as feeding patches (Fig. 1). Each patch was divided into four quadrants by a low, wooden wall (12 mm high), creating a total of 24 bait positions (minimizing cross-contamination between baits). Patches were designated 1–6, and quadrants A–D, for data handling purposes. A water bowl was sunk into the ground at the centre of the site for three reasons: to deter badgers from leaving the experiment in search of water; to prevent thirst-associated conditioning; and to prevent thirst from influencing the response of badgers to the repellents (Nicolaus et al. 1989b). For 6 weeks prior to the experiment, feeding patches were pre-baited using untreated baits. Observations confirmed that six badgers could feed simultaneously (one per feeder), without obvious conspecific interference. Just before dusk on each evening of the experiment, a 20-g pile of bait was placed in the centre of each of the 24 quadrants. These were treated or untreated as described below. The small bait quantities encouraged animals to sample all treatments. Bait and equipment were handled using disposable latex gloves.

Figure 1.

Patch layout. (a) Bait presentation on pre-trial and control nights; (b) bait presentation on treatment nights. 1–6, patches 1–6; A–D, patch quadrants; W, water bowl; R1–R3, baits treated with cinnamamide, capsaicin or ziram; C, untreated control bait. Together, (a) and (b) illustrate alternation between treatment and control nights in the trial phase.

The experiment lasted 124 nights, comprising a 68-night pre-trial phase followed by a 56-night trial phase.

Pre-trial phase

Each patch offered a choice of four piles of untreated bait (Fig. 1a).

Trial phase

Treatment (T) and control (C) nights were alternated (Table 1). On treatment nights (T1, T2, T3, etc.), each patch offered a choice of baits treated with each of the three repellents and a pile of untreated control bait (Fig. 1b). On control nights (C1, C2, C3, etc.), each patch offered a choice of four piles of untreated bait, as in the pre-trial (Fig. 1a).

Table 1.  Trial sequence. Treatment and control nights were alternated during the trial phase
PhaseNight typeNightsBaitsTreatment phaseTreatment phase nightsFeeder orientation
Pre-trialPre-trial1–68 (all nights)Untreated
TrialTreatment (T)69–123 (odd nights, T1–T28)Three repellents and untreated controlTPI TPII69–99 (odd nights, T1–T16) 101–123 (odd nights, T17–T28)Predictable Random
Alternated with control (C)70–124 (even nights, C1–C28)Untreated

The trial phase consisted of two parts. During part 1 (TPI), the position of treatments within each patch did not change; in the second (TPII), positions were randomized (Table 1). Inclusion of both predictable and random patch orientations allowed us to check that badgers were not ‘predicting’ the positions of treatments. We alternated control and treatment nights to check for evidence that feeding behaviour was subject to seasonal effects. Lack of seasonality would mean that treatment night responses could be compared legitimately with those from the pre-trial. Any predictability that this introduced should not have affected the relative way in which badgers responded to the different bait types presented on treatment nights.

behavioural event recording

Video data were successfully gathered on 111 of the 124 nights (pre-trial 56/68, treatment 27/28, control 28/28). We analysed behavioural information from every treatment night, and a sample of pre-trial and control nights (pre-trial n = 30, treatment n = 27, control n = 15). Alternate tapes were analysed from the pre-trial and control nights. Each tape was analysed continuously from bait presentation, until the point at which feeding at the trial site ceased.

An individual badger's behaviour was recorded as a series of events, each associated with a patch quadrant (1A−6D; Fig. 1a). Data for unmarked badgers were recorded separately. Events comprised the following: sample, bait/quadrant sniffed (nose over quadrant) or tasted, but eating (chewing) not observed; eat part, badger fed at the quadrant but stopped before the bait pile was gone; eat all, bait pile was completely finished (regardless of whether it had been started by another or the same individual during a previous event). We recorded the start time for each event, and an event ended when sampling or feeding stopped at the quadrant.

To prevent social interactions from affecting our measurement of repellent efficacy, we eliminated all events that were deemed to have terminated early because of social interaction (see the Appendix for definitions). On treatment nights, we recorded first-choice responses separately for each bait type. These consisted of all events (sample, eat part and eat all) occurring at a patch that offered a choice of four previously untouched baits. For each bait type, we calculated the percentage of first-choice responses during which the bait was finished (eat all event occurred).

We designated all sample and eat part events after which the badger moved directly to a different patch, as bait patch rejection events. Any bait type remaining uneaten on the patch at the end of such an event was allocated a point; for example, if a badger left a patch on which uneaten ziram and capsaicin baits remained, both ziram and capsaicin received a point. Points were tallied for each bait type on each night. To weight the scores for untreated baits, on pre-trial and control nights (when four piles of untreated baits were presented on each patch), we divided their tallies by four. This allowed direct comparison of weighted pre-trial and control night tallies (for untreated baits), with unweighted treatment night tallies (for ziram, capsaicin, cinnamamide and untreated control baits). Bait patch rejection was further weighted by (100/the number of visits made by the individual concerned) to account for badgers making different numbers of visits on the three types of night.

bait consumption

We collected, counted and weighed all uneaten pellets. This allowed us to calculate the percentage of baits consumed, while taking into account the effects of temperature and humidity on pellet weight. Provided all bait was eaten, bait survival time was recorded as the time between bait presentation and the final eat all event for that night. For any night on which baits remained uneaten 24 h later, bait survival time was recorded as 24 h (minimum value).

Next, we assessed whether there was a positional preference among badgers for certain patch quadrants, to identify any positional biases. For each night of the trials, the 24 quadrants were ranked 1–24, according to the order in which badgers finished the bait on each (eat all event occurred on the quadrant); for example rank 1 was assigned to the quadrant finished first. Joint ranks were allocated where the order in which baits were eaten was unclear. Then we assessed badgers’ preferences for the four bait types. For treatment nights only, each of the 24 bait piles was similarly assigned a rank, according to the order in which they were completely consumed. Again, joint ranks were allocated, where appropriate, and the six ranks then summed for each of the four bait types on each treatment night.

statistical analysis

Our observations were based on badgers visiting a single site. We emphasized effect sizes, rather than hypothesis testing (Oksanen 2001), as is appropriate for large-scale experiments where replication is limited. Non-parametric statistical tests were applied using SAS software (SAS Institute Inc. 1989). Using individual badgers as replicates, we used ranked paired (matched) comparisons (SAS proc freq; Cochran–Mantel–Haenszel, CMH, approximation to Friedman's χ2 test; SAS Institute Inc. 1989), to test for differences in badgers’ first-choice responses, and the amount of bait patch rejection directed at the four treatments. Where we identified a difference between treatments, we made pair-wise comparisons using Wilcoxon's sign rank test. On treatment nights, we tested the null hypotheses that the rank order of bait survival did not depend on bait position or treatment (CMH approximation to Friedman's χ2 test).



Seven marked badgers visited the feeding site, with four attending on between 81% and 93% of treatment nights. The mean number of marked badgers visiting each night was: pre-trial, 3·5 badgers (SE = 0·2); control, 3·1 (SE = 0·3); treatment, 3·9 (SE = 0·2). An unmarked individual(s) visited on 13 nights during the pre-trial, on two treatment nights, and never on control nights.

behavioural responses to repellents

First-choice responses

We analysed first-choice responses to assess whether individuals responded similarly to a fresh choice of the four treatments. Using individual badgers as replicates (n = 4), the percentage of responses where badgers ate all of the bait pile differed significantly across the four bait types (Friedman's χ2 approximation: inline image, P = 0·035; Fig. 2). There was a marginally significant difference between control and ziram baits, and between control and capsaicin baits (Wilcoxon's sign rank statistic: WSRS[1] = 4·0, P = 0·047; WSRS[1] = 3·7, P = 0·055). No other pairwise comparisons were significant (WSRS[1] ≤ 2·0, P ≥ 0·162). The pattern of behaviour was almost identical for three animals, with untreated control baits being eaten entirely (eat all) on 77–93% of first-choice occasions, and repellent-treated baits being only sampled on 94–100% of occasions. The fourth badger entirely consumed control baits on 57% of first-choice occasions, capsaicin on 14% and cinnamamide on 85%. No first-choice responses directed towards ziram involved eating: ziram-treated baits were only sampled.

Figure 2.

Median percentage of first-choice responses, made towards baits on treatment nights, which were ‘eat all’ (n = 4 badgers). Circles indicate values for individuals; *2, *3 and *4 indicate two, three and four overlaid points, respectively, of value zero; Con, untreated control; Cap, capsaicin; Cin, cinnamamide; Zir, ziram. Responses included all events occurring at a patch that offered a choice of four previously untouched baits. For each of the four bait types, we tallied separately the number of sample, eat part and eat all events. Bait types denoted by the same symbol did not differ significantly (WSRS, P  0·05).

bait patch rejection

Bait patch rejection tallies (patch rejection) indicated a particular repellent's ‘power’ to move the badger to a different feeding patch altogether, once an individual had sampled a bait. Using individual badgers as replicates (n = 5), patch rejection associated with untreated baits did not differ significantly across pre-trial, control and treatment nights (inline image, P = 0·472). However, rejection differed significantly across the four bait types presented on treatment nights (inline image, P = 0·004; Fig. 3). There was a significant difference in patch rejection for each pair of treatments (WSRS[1] ≥ 4·5, P ≤ 0·035), except for capsaicin and cinnamamide (WSRS[1] ≤ 0·2, P = 0·655). Median rejection associated with each of the three repellents was 20–85 times higher than that associated with untreated baits. Patch rejection associated with each repellent rose sharply to a peak around the seventh treatment night (thirteenth from first exposure), before returning to previous levels (for ziram peak see Fig. 4). Behaviour of known individuals closely reflected this pattern.

Figure 3.

Median bait patch rejection associated with each bait type on treatment nights (n = 5 badgers). Circles indicate values for individuals; *2 indicates two overlaid points; Con, untreated control; Cap, capsaicin; Cin, cinnamamide; Zir, ziram. Bait types denoted by the same symbol did not differ significantly (WSRS, P  0·05).

Figure 4.

Ziram bait consumption, and total bait patch rejection associated with ziram, on each treatment night, for all badgers. Percentage of bait eaten: black squares indicate values based on video evidence; white squares indicate estimates calculated from retrieved pellets. On T7*, all baits were eaten by unmarked badger(s). Black circles indicate ziram bait patch rejection; white circles indicate the minimum value for incomplete nights. Bait patch rejection data were included to illustrate the association between the decline in ziram baits eaten and the peak in ziram bait patch rejection.

bait consumption

All untreated baits were eaten on each night of the pre-trial phase (median bait survival time 24 min, Inter-quartile range, IQR, 20–35 min, n = 30 nights) and all control nights (median bait survival time 22 min, IQR 17–29 min, n = 13). All treated and untreated baits were eaten on the first and second treatment nights. All untreated baits, and baits treated with capsaicin or cinnamamide, continued to be eaten throughout the trial. However, from the third treatment night (fifth from first exposure), badgers began to leave ziram-treated baits uneaten. Badgers ate a total of 26·5% of ziram-treated bait over the whole trial phase.

The consumption of ziram-treated bait fell to zero at T9 (seventeenth from first exposure) and, thereafter, badgers tasted ziram only rarely. Individual patterns of sampling differed, but badgers tended to eat or sample ziram-treated baits on between two and six nights (spaced at varying intervals) towards the beginning of the treatment phase, before ceasing to sample or eat them altogether. On treatment night 7, the last night on which all ziram baits were eaten, this was entirely attributable to an unmarked badger(s) (Fig. 4). Small quantities were eaten by one of three individuals on T9, T10, T11 and T16. The four regular trial phase participants avoided ziram baits without sampling for the last 17, 18, 12 and 22, respectively, of the 28 treatment nights. Ziram baits remained completely untouched for the last 12 treatment nights (video footage was lost shortly before the end of treatment nights 22 and 23, and for the entirety of treatment night 13; see estimates on Fig. 4).

The decrease in consumption of ziram-treated bait (T5–T9; Fig. 4) coincided with a period during which regular visitors exhibited an increasing number of patch rejection events towards treated baits. Rejections reached sharp peaks around T7 (for ziram peak see Fig. 4). After this, rejections quickly returned to (and remained at) the lower levels observed towards the beginning of the trial phase, while ziram bait consumption remained negligible.

An analysis was performed to determine badgers’ preference order (if any) for the four bait types. First, we checked for any positional preferences among badgers for certain patch quadrants. In the pre-trial, there was a significant difference in the rank order from which different bait quadrants were eaten (inline image, P = 0·003, n = 30 nights). Badgers generally approached the site from one direction (from the bottom of Fig. 1). The positional bias observed in the pre-trial reflected the fact that, in this phase, they tended to eat from feeders in the order in which they were encountered (the increasing median rank, listed here for each feeder, illustrates the order in which badgers ate from them in this phase: feeder 3, rank 9 (IQR 4–16·5); feeder 4, rank 10 (IQR 6–16); feeder 2, rank 10 (IQR 6–18); feeder 5, rank 14 (IQR 7–18·5); feeder 1, rank 16 (IQR 10–20); feeder 6, rank 16 (IQR 8–20·5). When repellents were introduced, there was no longer a significant difference in the rank order from which quadrants were eaten (treatment nights, inline image, P = 0·999, n = 27; control nights, inline image, P = 0·459, n = 15), so bait position did not have a confounding effect in the trial phase.

The rank order in which treatments were taken differed within treatment nights (TPI, inline image, P < 0·001, n = 15 nights; TPII, inline image, P < 0·001, n = 12). Untreated controls were eaten before each of the three types of repellent-treated bait on all but the first treatment night (Fig. 5). Capsaicin and cinnamamide baits were eaten consistently later than controls, but their relative ranks were not distinguishable. Ziram was eaten last, or left uneaten, on all treatment nights.

Figure 5.

Summed rank of the order in which bait piles, of the four bait types, were eaten on each treatment night (eat all event occurred). This illustrates the order in which baits were eaten, with lower ranks representing bait types taken earlier, and higher ranks representing those taken later. Control, white squares; capsaicin, white triangles; cinnamamide, black squares; ziram, black triangles).


We tested the hypothesis that known individual badgers would demonstrate relative preferences for four bait types (three repellent-treated and one untreated). We established that wild mammals could develop CTA to a food-based repellent and then discriminate efficiently between treatments, thereby avoiding aversive baits without sampling. Once acquired, the aversion was maintained without additional reinforcement for the remaining 12–22 treatment nights (24–44 trial nights). This has implications for the protection of untreated foods from wild animals. Detailed behavioural analysis allowed us to observe the exact pattern of events taking place, and so deduce that the mechanism most likely involved was second-order conditioning. We believe that the results presented here already point to circumstances where such methods can be useful to wildlife managers.

We tested repellents together to assess their relative effects in a cafeteria-style experiment, adapting the methodologies of Andelt, Burnham & Manning (1991) and Andelt, Burnham & Baker (1994). This approach allowed comparison of individual badgers’ responses to the three repellents, while avoiding potentially confounding effects arising from prior experience, seasonal changes in natural food availability, behaviour or motivation to feed (Rogers 1974; Nolte & Barnett 2000). As a result, we are cautious in our conclusions regarding the likely effect of the compounds in isolation. Nevertheless, the effects of ziram were so clear here that it is difficult to believe that they would be less so in the absence of the other treatments. Furthermore, the badgers’ ability to discriminate precisely between the various treatments in our study is of practical predictive value, because free-ranging omnivores rarely spend a night consuming a single food (Macdonald & Barrett 1995). This suggests that the CTA, which badgers appeared to develop here, could allow them to avoid target food while continuing to eat alternative foods (Dimmick & Nicolaus 1990) and maintaining their territorial role (Nicolaus & Nellis 1987).

Ours was a conservative test of repellency, because the 68-night pre-trial phase probably created a significant expectation that foods at the site would be palatable. Also, repeated sampling of ‘safe’ untreated baits in the pre-trial may have endowed treated baits with some degree of learned safety, a phenomenon that has been shown to attenuate the development of CTA (Kalat & Rozin 1973). In addition, the majority of Wytham badgers do not survive beyond their fourth year (Macdonald & Newman 2002) and our sample was biased towards older animals (three of the four most regular visitors were between 5 and 13 years old), with extreme toothwear and in poor body condition. Such animals, if less able to forage naturally, may be more likely to become pests, so providing a good model for this research.

There was little between-individual variation in treatment responses. The consistency of bait patch rejection and other behaviour towards untreated baits, on pre-trial, treatment and control nights, indicated that there were no seasonal effects and that badgers had not formed general aversions to the site or to untreated baits. One possible exception might have been badger 70, which stopped attending after four treatment nights.

capsaicin and cinnamamide

Badgers appeared not to discriminate between capsaicin, cinnamamide and untreated baits on the first treatment night (Fig. 5), but thereafter formed a clear preference for untreated baits, followed by cinnamamide and capsaicin in no particular order, and then ziram. This preference was apparent in diverse behavioural measures (first-choice responses, bait patch rejection and order eaten), but would not have been detected by measuring bait consumption alone.

Badgers ate capsaicin, cinnamamide and untreated baits with equal eagerness on the first treatment night, which suggests that neither capsaicin nor cinnamamide, was immediately aversive, although capsaicin is considered to be innately repellent to mammals (Mason et al. 1991). Neither repellent proved promising as a feeding deterrent for badgers at the concentrations tested. Ultimately, we could not discriminate between the badgers’ treatment of capsaicin and cinnamamide, so while the reasons may have been different, the general measure of aversion to these two substances was similar.


Badgers preferred all other baits over those treated with ziram. On the first and second treatment nights, they finished ziram baits last, suggesting some innate avoidance of the taste (or irritancy) of ziram (Merck 1989). We used ziram incorporated with a ‘sticking agent’, a commercially available repellent formulation marketed as AAProtect™ (CAB International & the British Crop Protection Council 2001). AAProtect™ has an odour, acquired during manufacture, that is detectable to humans; this wanes as the formulate dries. If badgers were deterred at all by this odour, this was not sufficient to prevent them from approaching and eating ziram baits initially. Between T3 and T9 (fifth and seventeenth nights from first exposure), badgers left an increasing proportion of ziram baits uneaten. This decline in ziram consumption coincided with a sharp increase in the patch rejection behaviour associated with each repellent. Patch rejection peaked around the seventh treatment night (thirteenth from first exposure) before returning to previous levels (Fig. 4). Ziram bait consumption was practically zero over the last 20 treatment nights (40 trial nights). Individuals avoided ziram baits at a distance for the last 12–22 treatment nights (24–44 trial nights), and sampled them only very rarely. After eating all other baits, badgers often returned to walk around at the site, apparently ignoring the intact ziram baits (similar behaviour has been observed among racoons with CTA towards eggs; Semel & Nicolaus 1992). Our badgers stopped eating baits abruptly following full consumption, a feature characteristic of CTA (Dimmick & Nicolaus 1990). Under these conditions, it appears that ziram generated CTA in badgers.

Each of the regular visitors in this study demonstrated an individual pattern of sampling, and individuals stopped sampling ziram at different stages of the trial phase. This suggests that there was no significant social transmission of CTA. Although social communication about foods has not been studied in badgers, research on racoons and Norway rats Rattus norvegicus indicates that CTA is not socially transmitted to naive individuals (Semel & Nicolaus 1992; Galef 1997).

Taste is generally more likely to become associated with illness than other cues (Garcia & Hankins 1977), because taste and visceral information are processed in the same part of the brain (nucleus solitarius) whereas that relating to odour, and other non-taste cues, is not (Garcia, Clarke & Hankins 1973). When an undetectable agent causes CTA, the animal subconsciously associates the taste of the food with the illness produced, and subsequently avoids the taste of the food as a result (Garcia, Kimeldorf & Koelling 1955). Our badgers discriminated between baits according to treatment, indicating that they were able to detect ziram, and consequently developed an aversion to the taste of ziram (probably through post-ingestional effects), rather than to the taste of the baits themselves (Cowan, Reynolds & Gill 2000). Indeed, many naturally occurring toxins that produce CTA are highly flavoured (Nicolaus 1987) and may function in this manner.

When CTA develops, either to the taste of food or the taste of a detectable repellent, the target animal would have to bite or lick the food each time before being averted (Gustavson et al. 1974, 1976). This might necessitate the food becoming damaged (or killed) (Gustavson et al. 1976; Nicolaus & Nellis 1987). Avoidance without sampling is therefore key to protecting untreated foods; in fact, extinction of aversions is usually a result of sampling (Testa & Ternes 1977). Nicolaus and co-authors, have performed a number of studies testing CTAs as wildlife management tools for the protection of untreated foods from damage by free-ranging racoons, mongooses Herpestes auropunctatus and guilds of mammalian predators including American badgers Taxidea taxus. In these studies, animals avoided untreated foods at a distance, probably through second-order conditioning or potentiation (Nicolaus, Hoffman & Gustavson 1982; Nicolaus & Nellis 1987; Nicolaus et al. 1989c; Semel & Nicolaus 1992). Second-order conditioning is a two-stage process: first, the taste of food becomes aversive when paired with illness; secondly, non-taste cues become associated with the now aversive taste, and themselves inhibit approach or attack (Gustavson et al. 1974). Potentiation is a one-stage process whereby a weak cue for illness, for example odour or colour, can be facilitated to become a strong cue (Garcia & Rusiniak 1980; Westbrook, Clarke & Provost 1980).

Up to the night on which badgers stopped eating ziram baits, our behavioural analyses revealed an increase in patch rejection behaviour towards repellent-treated baits, and in particular ziram (Fig. 4). We conclude that, despite simultaneous presentation of four bait types, badgers developed CTA to the taste of ziram baits, and then second-order conditioning took place over this period of increased sampling/rejection. Badgers then avoided ziram baits, without further direct sampling, using odour as the aversive cue. The important question is whether they would have continued to avoid untreated baits that produced the same cues (Nicolaus et al. 1983). Such a phenomenon might be used to protect untreated foods that are destined for human consumption (e.g. crops, grain, eggs and fruit; Gustavson 1977; Nicolaus & Nellis 1987; Nicolaus 1987; Nicolaus et al. 1989a; Semel & Nicolaus 1992). However, the situation is not straightforward.

There are two ways in which CTA to an odour cue might theoretically be used to protect untreated foods from sampling: (i) a CTA agent, undetectable or detectable, could be used to create an aversion to the taste (of the food or the agent, respectively) and odour of the target food, such that conditioned animals subsequently avoided the odour of the food itself; or (ii) a bi-sensory aversive agent [combining a CTA agent (undetectable or detectable) with a novel, benign and effusive odour] could be used to create an aversion to the taste (of the food or the agent, respectively) and the odour cue provided, with the aim that target animals would subsequently avoid the odour cue. The odour might then be used to reduce feeding damage in sensitive areas.

It would seem preferable to use an undetectable CTA agent. The potential drawback of using an agent with a detectable taste, such as ziram, is that any olfactory barrier (whether natural food odour or added odour cue) may be breached, and the untreated food will taste different from the ziram-treated food that caused the original illness. However, no CTA agents that are undetectable and sufficiently safe have yet been registered for use. Even oral oestrogen, a uniquely successful undetectable CTA agent, may act as an abortifacient or teratogenic (Reynolds 1999). It is extremely unlikely ever to be registered as a repellent (or for any environmental use) in the UK. Nevertheless, despite the potential drawback of using a detectable aversive agent, our results demonstrated that badgers did not breach the odour barrier to sample ziram baits for the last 12–22 treatment nights (24–44 trial nights). This may well have been sustained if the experiment had continued. We conclude that: (i) badgers can learn to avoid foods on the basis of odour without sampling; and (ii) while the search continues for safe, undetectable CTA agents, we should further investigate the use of ziram (Baker et al., in press a, in press b ). The next step is to develop appropriate management strategies through field trials.

CTA particularly lends itself to the protection of foods that are vulnerable for fixed, predictable periods, for example crops susceptible to badger damage for a few weeks before harvest. Protection would therefore be required for this limited period only (Wilson 1993; Moore et al. 1999). We propose that a bi-sensory agent combining ziram with a novel, effusive odour cue, for example clove oil, might be used for this purpose (Baker et al. in press b). An irritating, or unpleasant-smelling, odour would prove undesirable and impractical (for humans and non-target species) for use in real-life wildlife management situations, whereas a novel, effusive odour should be readily associable and unlikely to have previously acquired connotations. For example, in order to protect a maize crop from badger damage, maize cobs treated with the bi-sensory agent could be dispersed on the ground around the growing crop and at the local badger setts. Treatment should take place prior to the predicted time of damage (Avery & Decker 1994), to prevent cobs acquiring learned safety through badgers sampling the ‘safe’ untreated ripening crop before conditioning begins and to allow sufficient time for the aversion to develop, i.e. for the learning curve to take place. The clove odour (alone) and sacrificial cobs (treated with the bi-sensory repellent) would be replenished throughout the sensitive period to reinforce the aversion.


This work was funded by DEFRA as part of a collaboration between the Wildlife Conservation Research Unit (WildCRU), the Central Science Laboratory (CSL), the Forestry Authority and the Agricultural Development and Advisory Service (ADAS). S. Baker. was supported by a grant from the Royal Society for the Prevention of Cruelty to Animals during preparation of the manuscript. Trials were performed under PSD Administrative Experimental Approval. We are grateful to colleagues at CSL: Allan Nadian formulated the capsaicin and cinnamamide; Jo Gurney assisted with bait treatment; Roger Quy commented on an earlier draft. Our thanks also to colleagues at WildCRU: Paul Johnson helped greatly with statistical advice, and provided useful comments; others joined us in trapping and marking badgers, work conducted routinely (under Home Office PPL 30/1216) through generous sponsorship by the People's Trust for Endangered Species. Thanks also to anonymous referees, for their insightful comments.