Grazing in heterogeneous environments: infra- and supra-parasite distributions determine herbivore grazing decisions

Transcription

1 Oecologia (2002) 132: DOI /s z BEHAVIOURAL ECOLOGY Michael R. Hutchings Iain J. Gordon Ilias Kyriazakis Ewen Robertson Frank Jackson Grazing in heterogeneous environments: infra- and supra-parasite distributions determine herbivore grazing decisions Received: 12 March 2002 / Accepted: 29 April 2002 / Published online: 2 July 2002 Springer-Verlag 2002 Abstract We tested the hypothesis that the infra-gastrointestinal parasite population of herbivores affects their grazing behaviour in relation to the supra-parasite population of parasites in the environment. Our first objective was to create a naturally heterogeneous sward structure of gaps and tussocks using a continuous grazing scheme. We then demonstrate that a nutrition vs. parasitism grazing trade-off occurs within that sward structure and that infra-gastrointestinal parasite populations affect the grazing decisions of herbivores faced with the trade-off. A pool of 50 naturally parasitised female Soay sheep and their lambs were used to create a heterogeneous tall, faeces-contaminated tussock/short, non-contaminated gap sward structure in a 1-ha experimental plot. Tussocks offered approximately 1.5 times greater forage intake but contained 5.5 times the number of strongyle parasites compared to the gaps. Following a 10-week period in which the heterogeneous sward structure was created, two 5-day periods of observations of sward structure selection (i.e. gap vs. tussock) were carried out. Twenty female Soay lambs were divided into two groups of ten (balanced for live-weight) immediately prior to the start of the observation period. One of the groups of lambs was treated with an anthelmintic drench before the start of the second observation period creating two levels of parasitism (high and low). On each observation day 5-min focal observations were carried out on each animal at least twice a day, during which time the number of bites taken from gaps and tussocks were recorded along with the number of steps. During the first period of observations, all animals rejected the relatively tall, M.R. Hutchings ( ) I. Kyriazakis Animal Nutrition and Health Department, Scottish Agricultural College, West Mains Road, Edinburgh, EH9 3JG, UK m.hutchings@ed.sac.ac.uk Tel.: , Fax: I.J. Gordon E. Robertson Macaulay Institute, Craigiebuckler, Aberdeen, UK F. Jackson Moredun Research Institute, Pentland Science Park, Penicuik, UK faeces-contaminated tussocks for grazing to a similar extent and had similar bite and step rates. During the second period of observations all animals showed reduced rejection of the tussocks relative to the first week, however, animals with a reduced parasite population showed a greater reduction in rejection as compare to the highly parasitised animals. We conclude that the infraand supra-distributions of parasites within herbivore hosts and the environment greatly impact on herbivore grazing behaviour and foraging decisions and thus the structure and heterogeneity of grazed ecosystems. Keywords Foraging trade-off Herbivore Parasite burden Parasitism Physiological state Introduction The eggs of many species of gastrointestinal parasites, which constitute the most pervasive challenge to mammalian herbivore hosts (Coop et al. 1982; Gulland 1992), develop in faeces deposited by herbivores on vegetation and migrate onto the surrounding sward as infective third-stage larvae (Sykes 1987). Herbivores cannot detect the parasites themselves on swards and thus use the presence of faeces as an environmental cue to the presence of parasites (Cooper et al. 2000). Here we test the overall hypotheses that supra-parasite populations through their association with faeces are a major source of heterogeneity of forage resources and, and that infraparasite populations of herbivores affect the host s grazing decisions in this heterogeneous environment. Herbivores avoid grazing swards contaminated with faeces and thus parasites: fresh faeces are avoided most strongly and this avoidance declines with the age of faeces (Hutchings et al. 1998). It takes a few weeks for the eggs of gastrointestinal parasites (e.g. approximately 3 weeks in the case of Ostertagia circumcincta) in faeces to develop into infective third-stage larvae. During the initial period of faecal avoidance, when parasites develop into infective-stage larvae, the surrounding sward

2 454 may become relatively tall, creating patches of tall sward contaminated with faeces and parasites. This faecesavoidance behaviour may then be seen as leading to a mosaic of tall and short sward patches and a trade-off between the benefits of grazing tall swards (i.e. increased herbage intake rate) (Penning et al. 1991; Gordon et al. 1996) and the costs associated with parasite ingestion. The grazing decisions of herbivores faced with such trade-offs will determine both their nutrient and parasite intake. Sward height and level of faeces contamination may enable grazing mammals to assess the relative costs and benefits of a grazing trade-off, which in turn determines their grazing behaviour (Lafferty 1992). However, the relative costs and benefits of a trade-off may also be affected by animal physiological state (Lozano 1991; Coop and Kyriazakis 1999). The increased penalties associated with parasite intake (i.e. endogenous loss of nitrogen and damage to gastrointestinal tissue) in animals with an existing parasite burden (Poppi et al. 1986; Brown et al. 1991), may increase faecal avoidance behaviours relative to parasite-immune or parasite-naïve animals (Hutchings et al. 1999, 2000, 2001a). However, this effect of parasitic state on host foraging behaviour may be due to differences in immune state or differences in size of the infra-parasite population (parasite burden). For example, both parasite-naïve and parasite-immune animals have no parasite burden but they differ in their aversion to grazing faeces-contaminated swards, suggesting that immune state may affect herbivore grazing decisions in relation to supra-parasite populations in the environment (Hutchings et al. 1999, 2001a). The role of infra-parasite population size in determining herbivore foraging behaviour is less clear. In mammalian herbivore hosts the majority of fitness losses, mainly through reduced growth rate, occur during the period of acquisition and expression of immunity to gastrointestinal parasites, and the level of loss is positively related to the level of parasite intake and size of parasite burden (Coop et al. 1982). It might then be expected that herbivores would act to minimise the fitness losses to gastrointestinal parasites by avoiding faeces- and thus highly parasite-contaminated swards, and that faeces avoidance is increased with increasing parasite burden. The first aim of this experiment was to create a natural gap/tussock sward mosaic through continuous grazing by a herbivore species, e.g. similar to that found on St Kilda where feral Soay sheep (Ovis aries) graze a heterogeneous gap/tussock sward structure (Jewell et al. 1974). Secondly we aimed to demonstrate that the gap/tussock mosaic represented a nutrition vs. parasitism trade-off between grazing tall, parasite-contaminated tussock swards and short, non-contaminated gap swards. Finally we aimed to demonstrate that, all else being equal, parasite burden of herbivores affects their grazing behaviour in relation to the gap/tussock sward mosaic and thus the trade-off between nutrient and parasite intake. Specifically, we test the hypotheses that: (1) tall tussock swards in a natural gap/tussock sward mosaic offer increased forage and parasite intake relative to short, gap swards, and (2) the reduction of parasite burdens in Soay sheep results in an increased selection for tall, parasite-rich tussock swards when grazing a natural gap/tussock sward mosaic. Materials and methods Experimental design The experiment was conducted in August 1999 at the Macaulay Land Use Research Institute s Hartwood Research Station where a 1-ha site was allocated for the experimental field plot. The plot was established on a mixed grass/clover pasture (Perennial ryegrass, Lolium perenne and Kent Wild White, Trifolium repens, a low growing clover). The experimental timetable (total of 12 weeks) was divided into: (1) a 10-week period during which the short gap/tall tussock sward structure was created (weeks 1 10), and (2) a subsequent 2-week behavioural observation period (weeks 11 and 12) (that was subdivided into two 5-day periods (days 1 5 in week 11 and 8 12 in week 12). The experimental design was based on: (1) the use of a pool of 50 Soay ewes and their lambs managed to create a gap/tussock sward mosaic, and (2) two levels of parasite burden (high=control and low=anthelmintic treatment) in animals grazing the natural gap/tussock mosaic (gap=relatively short, noncontaminated swards and tussocks=relatively tall, faeces-contaminated swards). The anthelmithic treatment was imposed between the first and second week of behavioural observations and thus the interaction between the main effects of week and anthelmintic treatment was used to test the hypothesis that a reduction in parasite burden affected herbivore grazing decisions. Animals Soay sheep are a primitive breed of sheep and several feral flocks persist in the UK, notably on the island of Hirta in the St Kilda archipelago where they persist in the absence of predators or competitors and graze a tall tussock/short gap sward mosaic (Jewell et al. 1974). A pool of 50 female Soay sheep and their 70 lambs (mean lambing date 17 March 1999) maintained under minimal parasite-management conditions were released into a 1-ha experimental plot of homogeneous sward height (approximately 7 cm) (week 1) and were managed to create a gap/tussock sward mosaic similar to that found under natural conditions (e.g. St Kilda) (Hutchings et al. 2002). The stocking rate was manipulated to minimise grazing of faeces-contaminated tall tussocks and maximise the sward surface height difference between the tall tussock patches and the short gap patches when creating the gap/tussock sward mosaic. Lambs were weaned (16 June 1999) by removing the dams from the experimental plot at the end of week 7. Twenty of the remaining female (ewe) lambs were allocated to one of two treatment groups (high and low parasite burden), balanced for live-weight at the start of week 10. The remaining animals were removed from the experimental trial, reverting back to their normal minimal management grazing pastures. Experimental treatment of the low parasite burden group occurred before the second week of observations when the animals were treated with a broad-spectrum anthelmintic (Systamex; 2.5 ml 10 kg 1 live weight) to reduce their gastrointestinal parasite burdens. It was known from ongoing monitoring of drench resistance by parasites that the nematode parasite population at the Hartwood Research Centre showed some resistance to white drenches such as Systamex (F. Jackson, personal communication). It was then expected that the dosing of sheep with a white drench would result in a reduction, but not a complete removal, of their parasite burden. The experiment was designed to be a severe test of the hypothesis that level of gastrointestinal parasite burden of hosts affects their grazing decisions in relation to the supra-parasite distributions in the environment. It was also expected that no differences in immune state would be apparent between ani-

3 455 mals in the two parasite treatment groups. The immune state of a herbivore is, in part, stimulated by the ingestion of infective-stage larvae (Gasbarre 1997; Maizels et al. 1999). The incomplete removal of worm burden and the continuing ingestion of parasites by the anthelmintic-treated animals grazing the parasite-contaminated pasture in the second period of the behavioural observations (week 11) was expected not to affect their immune state within the time period of observations (i.e. 1 week) compared to the nontreated control animals (Gasbarre 1997; Maizels et al. 1999). The experimental manipulation of worm burdens could, therefore, be seen as creating two parasite burdens whilst maintaining a consistent immune state across both anthelmintic treatments. The consistency of faeces was used to monitor all animals for signs of clinical disease (i.e. the onset of diarrhoea). To enable easy identification in the field, each sheep was marked with a combination of coloured tape on the horns and sprays on the rump. Swards Weekly stratified sward height measurements started in week 5 when 20 sward surface heights were taken using a sward stick (Barthram 1985) from both faeces-contaminated tussocks and non-contaminated gap swards whilst walking a W-transect. The behavioural observations started the week after the tussock sward height declined (i.e. week 10) (see Fig. 1). This indicated that the animals avoidance of the faeces-contaminated tussocks was declining (i.e. level of grazing tussocks had increased in some animals). Fig. 1 Mean (±SE) sward height of the gaps ( short, non-contaminated) and tussocks ( tall, faeces-contaminated). Day 0 represents the time animals were released into the experimental plots Measurements Faecal egg counts (FEC) were used as an indicator of parasitic infection of the sheep (Coop et al. 1982). Faecal samples were taken from the rectum of sheep and analysed for nematode eggs per gram of faeces (e.p.g.) using a modification of the flotation method described by Christie and Jackson (1982). The Soay sheep flock was not used to frequent handling and so this was minimised. Weekly faecal sampling of all animals started at the beginning of the period of behavioural observations (week 10) and continued through to the end of the experiment (week 12). Similarly, blood samples were taken once a week from the sheep by jugular venepuncture. The samples of blood serum were analysed for concentrations of pepsinogen, using a modification of the method described by Mylrea and Hotson (1969). Concentrations of pepsinogen are expressed as units of tyrosine, where 1 unit=1 µm tyrosine released per litre of plasma per minute at 37 C. Live weights of sheep and sward heights were taken at the start of the first week of behavioural observations (start week 10), the day of anthelmintic treatment of one of the animal groups (start week 11) and on the last day of the experiment (start week 12) (see Fig. 1). At the start of weeks 10, 11 and 12, three composite mean sward samples of gap and tussock vegetation were collected by clipping small areas of sward from across the field plot and analysed for dry matter (DM) content, nitrogen content using the method of Pella and Colombo (1973) and for in vitro organic matter digestibility (OMD) using the method of Alexander and McGowan (1966). This method ensured minimal sampling impact on sward structure and availability during the behavioural observation period. At the end of the experiment ten tussocks were chosen at random in the experimental field plot. The tussocks were clipped using electronic shears to the soil surface to create a total tussock sward sample. The gap vegetation surrounding the tussock was sampled by clipping herbage to the soil surface to create a gap sample. Sward nematode larval counts were then determined for each of the resultant 20 samples using a modification of the Weybridge method. The sward samples were washed in tap water with a few drops of detergent (Tween) (1 g material to 40 ml tap water). Following washing the herbage was removed and squeezed to remove the excess water. The herbage was then dried in a hot air oven to provide a DM estimate. The washings from the herbage were sedimented and the supernate removed with a vacuum line. The collected washings and larvae were then passed over a series of sieves (210 µm, 112 µm and 38 µm) to remove the coarse and very fine debris. Small amounts of the material collected from the 38-µm sieve were stained with helminthological iodine (allowing differentiation between parasitic and free-living nematode larvae) and examined directly at 100 magnification to determine the numbers of L 3 parasite larvae present in the sample. Species identification enabled "Nematodirus larvae to be distinguished from the remaining strongyles. Sward larval counts were expressed per unit herbage mass [per gram of DM (g DM)]. Five sward heights (using a sward stick) were taken from each tussock and gap stratum prior to cutting. Five hundred random pluck samples were taken from the remaining gap and tussock vegetation at the end of the experiment using thumb and forefinger to estimate the relative amount of herbage that would be consumed by the sheep when grazing the two sward types. The 500 pluck samples were pooled into composite means for gap and tussock, which were analysed for nitrogen content and in vitro OMD as above. Photography was used to determine the relative amounts of gap and tussock swards that made up the field plot. Photographs were taken from two observation towers (from which behavioural observations were made, see below) that overlooked two areas of the field plot. Photographs were taken from a fixed position, 5.2 m above ground level at the start of weeks 10, 11 and 12 (i.e. to coincide with behavioural observations see below). The camera (35 mm SLR camera) was fitted with a Zeiss Distagon 28 mm ƒ 2.8 lens (Carl Zeiss, Oberkochen, West Germany) and mounted on a tripod. Infra-red transparency film (Kodak EIR Ektachrome Professional Infrared; Eastman Kodak, Rochester, N.Y.) was exposed at ISO 250 using a lens aperture of ƒ 11, giving rise to a shutter speed of approximately 1/60 s. Two sets of monotone, but differently hued images (magenta and orange) were created by exposing the film with and without a yellow filter (Kodak Wratten K2; Eastman Kodak). The field was processed in E6 (Kodak) chemistry and images were scanned and stored on compact disc using a commercial drum scanner at a resolution of 6,000 dpi, resulting in individual image files of 18 mb each in which dark/light areas related to tussock/gap sward types. The weather was bright and sunny and the swards were dry on each day photographs were

4 456 taken (between 1300 and 1400 hours). To ensure that the same area was analysed at each time point for each tower, four white pegs were placed in the field such that the area of field captured by the camera was marked. The proportion of gap and tussock sward types in the scanned photographs were determined via a three-stage process: 1. Registration the pegs which marked the corners of each plot were used as control points for registering each image. The software used for the registration process was the geometric rectification module of ERDAS IMAGES (ERDAS 2001). 2. Classification Schowengerdt s (1983) supervised classification process was used to classify the gap and tussock sward types. 3. Class proportions the proportion of the different classes was determined by calculating the proportion of pixels present in each class (gap and tussock) in each image (Lillesand and Kiefer 1979). The mean of the results from the two different techniques (i.e. with and without the filter) was calculated prior to analysis. Animal grazing behaviour and sward patch selection Grazing behaviour was measured using direct observations to record both grazing behaviour and sward (gap/tussock) selection of the sheep. Observations were made from the towers described above. On each day of the experiment two observers recorded the behaviour of randomly selected individuals for 5 min during periods of activity. All individuals were recorded at least twice a day. A total of 525 five-min recording periods over 10 days were carried out on the 20 sheep. The following grazing variables were recorded in each 5-min observation: 1. Number of bites taken from the gap and tussock patches. A bite was defined as a head pull associated with the severing of herbage (Newman et al. 1992). 2. Bite rate (bites s 1 ). 3. Number of steps taken (steps s 1 ). A step was defined as the forward and/or backward motion of a chosen front leg. Statistical analyses ANOVA was used to determine how sward type (i.e. gap/tussock) and week (week 10, 11 or 12) affected the DM content (g kg 1 ), nitrogen content (g kg 1 DM) and in vitro OMD (g kg 1 DM) of swards. Residual maximum likelihood (REML) (Patterson and Thompson 1971) was used to estimate the mean values for grazing parameters [bite rate (bites s 1 ), step rate (steps s 1 )] and proportion of bites taken from tussocks. The GENSTAT REML (Lawes Agricultural Trust 1993) procedure was used which estimates SEs and SEs of the differences (SED) for the grazing parameters. The REML model included animal as a random effect and anthelmintic treatment and week of the behavioural observation and their interaction as fixed effects. It is possible that due to social facilitation the variances associated with step rates of the Soay sheep may be underestimated, although the effects are generally small (Rook 1998). t-tests were used to determine whether the estimated proportion of bites taken from tussock swards differed from the proportional availability of tussock swards. Wald tests from the REML routine were used to determine significant differences with the Wald statistic (w) quoted along with the relevant degrees of freedom and probability value for the fixed effects (Lawes Agricultural Trust 1993). FECs were log-transformed [log(x+1)] and proportions were arcsine transformed prior to their use in parametric statistical analyses (Zar 1984). Backtransformed FEC and proportion means are given in tables with upper and lower 95% confidence limits due to the restriction on backtransforming SEs (Zar 1984). Results Parasitic status of the sheep FEC of the animals at the start of the behavioural observations (week 10 i.e. allocation to treatment group) were similar across treatment groups [backtransformed mean FEC (95% confidence interval), high burden=548 e.p.g. ( ), low burden=509 e.p.g. ( )] (t 18 =0.38, P>0.5). No clinical signs of disease were observed during the first week of behavioural observations (i.e. no diarrhoea), which suggests that the animals had acquired a sub-clinical parasitism. At the time of implementation of parasite treatment (week 11, 1 day prior to anthelmintic treatment of the non-parasitised treatment group) FEC of all sheep were similar [high burden=346 e.p.g. ( ) and low burden=489 e.p.g. ( )] (t 18 =1.78, P>0.05). At the end of the experiment (week 12) the high burden animals had significantly higher FEC [407 e.p.g. ( )] than the low burden animals [109 e.p.g. (59 203)] (t 18 =3.54, P<0.01). At the time of allocation of sheep to treatment groups (week 10), and implementation of the anthelmintic treatment (week 11) serum concentrations of pepsinogen were similar across all animals [mean±se mu tyrosine: week 10 high burden=452±29 and low burden=203±173 (t 18 =1.34, P>0.1); week 11 high burden=756±230 and low burden=459±80] (t 18 =1.19, P>0.1). By the end of the experiment (week 12) concentrations of pepsinogen were significantly greater in high burden compared to the low burden sheep (mean concentrations of pepsinogen on week 12, mu tyrosine±se: high burden= 724±122; low burden=317±23) (t 18 =3.24, P<0.01). Live-weight of the animals did not differ across or within anthelmintic treatments during the period of behavioural observations [overall means live-weight±se of the 20 experimental sheep at the start of the behavioural observations (week 10), was 9.0±0.38 kg and at the end of the experiment (week 12) 9.1±0.48 kg]. Sward structure and composition At the start of the period of behavioural observations tussocks were approximately 4 times taller than gap swards (Fig. 1). Gap and tussock swards did not differ in their nitrogen or in vitro OMD throughout the experiment (Table 1). Negligible numbers of Nematodirus spp. larvae were recovered from the swards and, therefore, they were not statistically analysed. At the end of the experiment tussocks contained 3 times greater burdens of strongyle parasitic larvae than gap swards on a DM basis (Table 1). Composite mean pluck samples from tussocks contained marginally lower N (21.8 g kg 1 DM) but marginally greater OMD (744 g kg 1 DM) than from gap swards (23.4 g kg 1 DM and 715 g kg 1 DM, respectively). Five hundred pluck samples from gap and tussock swards produced 93.8 g and g of fresh matter

5 457 Table 1 Effects of sward type on sward characteristics. Values are means of nitrogen content and in vitro organic matter digestibility (OMD) [g kg 1 dry matter (DM)], and sward strongyle larval counts (SLC) (larvae kg 1 DM). SED SE of the difference, NS not significant (P>0.05) Sward Overall SED Effects mean Gap Tussock Week Sward Week sward Nitrogen Week NS NS NS Week Week OMD Week NS NS NS Week Week SLC Week ( ) 896 (642 1,249) 573 ** *P<0.05, **P<0.01 Table 2 Effect of parasite burden on the behaviour of Soay sheep grazing a heterogeneous sward. High burden (H)=natural sub-clinical gastrointestinal parasite infection and low burden (L) (anthelmintic treated at week 11)=control. Values given are means of bite rates, step rates and backtransformed mean proportion of bites from tussocks (with upper and lower 95% confidence limits in parentheses) Parasite Week 11 Week 12 Mean SED Effects burden H L H L Week Treatment Interaction Bite rate (bite s 1 ) NS NS * Step rate (step s 1 ) NS NS NS Proportion bites a a a a *** *** *** from tussocks ( ) ( ) ( ) ( ) *P<0.05, ***P<0.001 based on Wald statistics a Proportion of bites from tussocks was significantly less that the proportional tussock area of field plot (i.e. 0.45) (FM) and 24.9 g and 38.6 g of DM, respectively. Relative to gap swards and based on 500 pluck samples, tussocks contained 5.50 times greater numbers of strongyle parasite larvae per gram DM and offered some 1.44 times greater nitrogen per gram DM and 1.55 times greater DM intake rate. Tussock/gap availability The photographed areas of the field plot used to determine gap and tussock availability covered (Tower 1) (Tower 2)=0.231 ha (23% of the 1-ha field plot). The overall proportional area of tussock sward in the two photographed areas (±SE) was 0.45±0.030 (i.e. on an area basis, 45% of the field plot consisted of tussocks) and did not differ between the areas in front of the two towers (F 1,5 =9.74, P>0.05) nor across week of the experiment (F 2,5 =6.73, P>0.1). Sward selection and grazing behaviour There was a significant anthelmintic treatment effect, week effect and anthelmintic treatment week interaction in tussock selection (Table 2). When compared with tussock availability (i.e. 0.45) all animals significantly and strongly avoided tussocks for grazing in the first week of the grazing observations. The overall strong, and significant avoidance of tussocks was also apparent in the second week of behavioural observations. However, in the second week of behavioural observations all animals significantly reduced their avoidance of tussocks for grazing and the low parasite burden (anthelmintic treated) animals significantly more so than the high burden animals (Table 2). At the end of the experiment the low parasite burden animals rejected tussocks to a significantly lesser degree then the high parasite burden animals, however, all animals showed an overall avoidance to tussocks relative to their availability throughout the experiment. There was a significant week treatment interaction on animal bite rates. All animals had similar bite rates in the first week of behavioural observations. However, in the second week, low burden animals had lower bite rates than high burden (Table 2), probably because they were spending more time grazing the tussocks, which slow down bite rate. Anthelmintic treatment and week of the experiment had no effect on animal step rate nor was there an interaction (Table 2). Discussion The first objective of this study was to establish a heterogeneous gap/tussock sward structure similar to that found in natural systems (e.g. on St Kilda; Hutchings et al. 2002). It was expected that faeces-avoidance behaviour

6 458 by herbivores grazing a homogeneous sward would create a heterogeneous sward structure and a grazing tradeoff between nutrition and parasitism (Hutchings et al. 2001b). The non-contaminated, homogeneous (7 cm) sward of the experimental plot resulted in no trade-off being present at the start of the experiment. Five weeks after the Soay sheep were released into the plot there was a heterogeneous sward structure with the presence of distinctive gap (relatively short) and tussock (relatively tall) sward types. This heterogeneous sward structure was maintained throughout the experiment. The pluck samples taken at the end of the experiment from the gap and tussock sward types suggested that grazing the relatively tall tussock swards could yield some 1.5 times more DM and nitrogen intake than grazing the relatively short gap swards. However, all else being equal grazing tussocks could yield some 5.5 times on a per unit mass basis more strongyle parasite larvae than grazing the gap swards. Sheep then faced a trade-off between the advantages of grazing relatively tall faeces-contaminated swards offering greater forage and parasite intake and grazing relatively short less-contaminated swards. Faced with this trade-off, herbivores could be seen as facing a dilemma, in that their behavioural strategies to maximise nutrient intake (select tall swards i.e. tussocks) conflicted with their strategies to avoid parasites (avoid faeces i.e. tussocks). In this experiment the proportional area of tussock patches in the swards was 0.45; close to the maximum found on St Kilda (0.35) (Milner 1999; Hutchings et al. 2002). As the trade-off is created through the deposition of faeces on pasture the scale of spatial distribution of the trade-off (i.e. area and number of faeces-contaminated tussocks) in the environment is dependent on animal excretory behaviour. Soay sheep on St Kilda and sheep in agricultural pastures in general do not defecate at random but tend to concentrate faeces in either rest or shelter areas (Haynes and Williams 1993; Jewell et al. 1974). The distribution of faeces in the experimental plot would then be expected to be representative of both agricultural systems and more natural systems like St Kilda. The effect of the experimental parasite manipulations to determine the effect of infra-parasite populations on the grazing behaviour of hosts in relation to the supra-parasite populations in the environment might then be relevant to herbivores in both natural and agricultural systems. Faced with the trade-off all animals similarly avoided grazing the tussocks during the first week of the behavioural observations (week 10 of the experiment). This suggests that all animals detected the faeces in the tussocks and they preferred to graze the relatively short non-contaminated gap sward. The height of vegetative swards is a powerful cue used by herbivores to select for increased forage and hence nutrient and energy intake rate and is thought to play a key role in determining herbivore diet selection (Illius and Gordon 1990; Hutchings et al. 2000). The fact that all animals maintained a strong and significant selection for gap swards during the first week of behavioural observations suggests that the perceived costs of the trade-off in the first week of the experiment outweighed the benefits. During the second week of the behavioural observations, after half of the Soay sheep had their parasite burdens greatly reduced, level of tussock election was affected by anthelmintic treatment. All animals in the second week of behavioural observations showed increased election for tussocks compared to the first week. This is likely to be due to temporal effects on the relative costs and benefits of the trade-off perceived by the grazing animal. In vegetative swards the previous strong selection for gap swards and avoidance of tussocks is likely to increase the height difference between the two resulting in a reduction in the cost:benefit ratio of the trade-off. Also, the level of avoidance of the tussock is likely to diminish over time as the cue used by the sheep to avoid parasites (faeces) is reduced through the decomposition of the faeces, again resulting in a reduction in the cost:benefit ratio of the trade-off. It has previously been shown that faeces avoidance in sheep is a strong and inherent behaviour (Hutchings et al. 1998), but the attraction of sheep toward tall swards can completely overcome sheep avoidance of faeces (Hutchings et al. 2000). Both anthelmintic treatment groups increased their level of selection for tussocks in the second week of behavioural observations, however, the low parasite burden, anthelmintic-treated animals increased their election for tussocks to a significantly greater degree than the high burden animals. Tussocks, however, remained significantly avoided by all animals at the end of the experiment. This suggests that overall the perceived costs of the trade-off were greater than the perceived benefits throughout the experiment for both the high and low parasite burden animals, however, the perceived cost:benefit ratio was less for the low parasite burden animals. Any increased selection for the relatively tall tussocks by the low parasite burden animals in the second week of behavioural observations was expected to be associated with an increase in bite mass and thus a reduction in their bite rate, which is limited by rate of mastication (Black and Kenney 1984). The increase in election for tussocks by low parasite burden animals was associated with a reduction in bite rate and further supports our second hypothesis. The results of this experiment are consistent with the idea that variation in grazing behaviour may be due to parasite burden of the animal. The dosing of half of the Soay sheep with a fast-acting anthelmintic effectively created two groups of animals with the same immune status but with different parasite burdens (high and low). The acquisition and expression of immunity to gastrointestinal parasites is thought to be driven by ingested parasitic larvae (Gasbarre 1997; Maizels et al. 1999). The anthelmintic-treated Soay sheep would have continued ingesting parasite larvae during the second week of behavioural observations and thus their reduction in worm burden (but not complete removal of worm burden) was not expected to result in a change in immune state within the time period of the experiment. The reported changes in behaviour post anthelmintic treatment are thus likely

7 to be due to changes in worm burden rather than immune state. This is the first time that parasite burden alone (i.e. without the confounding immune state effect) has been shown to affect herbivore grazing decisions. The reduction in worm burden and thus challenge to the host s fitness by the parasites in the anthelmintic-treated sheep enabled them to take greater advantage of the nutritional benefits associated with the trade-off. The speed of the effect of worm burden reduction on herbivore host behaviour (i.e. days) reported in this experiment is consistent with the observations that parasite-induced anorexia in herbivores is lost almost immediately after removal of the parasite burden (see Kyriazakis et al. 1998). This suggests that herbivore hosts are sensitive to the presence and number of gastrointestinal parasites (i.e. infraparasite population) and alter their behaviour in response to levels of parasitism by balancing the costs and benefits of grazing decisions in relation to the distribution of parasites in the environment. This experiment has shown that selective grazing in relation to the distribution of parasites in the environment creates a highly heterogeneous sward structure similar to that found in natural and agricultural systems. The heterogeneous sward structure may contain a trade-off between the benefits of increased forage intake rate, through grazing relatively tall swards, and costs associated with parasite ingestion. The herbivores grazing behaviour in relation to this trade-off will determine their subsequent intake of nutrients and parasites and thus fitness and survival. Faced with this trade-off, herbivores trying to maximise fitness then face a dilemma in that their behavioural strategies to maximise nutrient intake (select tall swards) conflict with their strategies to avoid parasitism (i.e. faeces). The grazing behaviour of herbivores when faced with such a trade-off may depend on their parasite burden as it affects the relative costs and benefits associated with the nutrition vs. parasitism trade-off. We conclude that the infra- and supra-distributions of parasites within herbivore hosts and the environment greatly impact on herbivore grazing behaviour and foraging decisions and thus the structure and heterogeneity of grazed ecosystems. Acknowledgements The authors would like to thank Elizabeth Jackson and Dave McBean who gave parasitological technical support, Dave Allcroft of Biomathematics and Statistics Scotland for statistical support, David Miller for analysing the photographic images and the staff at Hartwood Research Station for management of the Soay flock. Thanks are also due to R. L. Coop and J. A. Milne for useful discussions and comments on previous drafts of this manuscript. The project was in part funded by the Natural Environment Research Council. The Scottish Agricultural College, Macaulay Land Use Research Institute and Moredun Research Institute receive support from the Scottish Executive, Environment and Rural Affairs Department. References Alexander RN, McGowan M (1966) The routine determination of in vitro digestibility of organic matter in forages an investigation of the problems associated with continuous large-scale operations. J Br Grassl Soc 21: Barthram GT (1985) Experimental techniques: the HFRO sward stick. Biennial report Hill Farming Research Organisation, Penicuik, pp Black JL, Kenney PA (1984) Factors affecting diet selection by sheep. II. Height and density of pasture. Aust J Agric Res 35: Bown MD, Poppi DP, Sykes AR (1991) The effect of post-ruminal infusion of protein or energy on the pathophysiology of Trichostrongylus colubriformis infection and body composition in lambs. Aust J Agric Res 42: Christie M, Jackson F (1982) Specific identification of strongyle eggs in small samples of sheep faeces. Res Vet Sci 32: Coop RL, Kyriazakis I (1999) Nutrition-parasite interaction. Vet Parasitol 84: Coop RL, Sykes AR, Angus KW (1982) The effect of three levels of intake of Ostertagia circumcincta larvae on growth rate, food intake and body composition of growing lambs. J Agric Sci 98: Cooper J, Gordon IJ, Pike AW (2000) Strategies for the avoidance of faeces by grazing sheep. Appl Anim Behav Sci 69:15 33 ERDAS (2001) ERDAS IMAGE on-line manual, com/software/image Gasbarre LC (1997) Effects of gastrointestinal nematode infection on the ruminant immune system. Vet Parasitol 72: Gordon IJ, Illius AW, Milne JD (1996) Sources of variation in the foraging efficiency of grazing ruminants. Funct Ecol 10: Gulland FMD (1992) The role of nematode parasites in Soay sheep (Ovis aries L.) mortality during a population crash. Parasitology 105: Haynes RJ, Williams PH (1993) Nutrient cycling and soil fertility in the grazed pasture ecosystem. Adv Agron 49: Hutchings MR, Kyriazakis I, Anderson DH, Gordon IJ, Coop RL (1998) Behavioural strategies used by parasitized and nonparasitized sheep to avoid ingestion of gastro-intestinal nematodes associated with faeces. Anim Sci 67: Hutchings MR, Kyriazakis I, Gordon IJ, Jackson F (1999) Tradeoffs between nutrient intake and faecal avoidance in herbivore foraging decisions: the effect of animal parasitic status, level of feeding motivation and sward nitrogen content. J Anim Ecol 68: Hutchings MR, Kyriazakis I, Papachristou TG, Gordon IJ, Jackson F (2000) The herbivores dilemma: trade-offs between nutrition and parasitism in herbivore foraging decisions. Oecologia 124: Hutchings MR, Kyriazakis I, Gordon IJ (2001a) Herbivore physiological state affects foraging trade-off decisions between nutrient intake and parasite avoidance. Ecology 82: Hutchings MR, Gordon IJ, Kyriazakis I, Jackson F (2001b) Sheep avoidance of faeces creates a trade-off between nutrition and parasitism in foraging decisions. Anim Behav 62: Hutchings MR, Milner JM, Gordon IJ, Kyriazakis I, Jackson F (2002) Grazing decisions of Soay sheep (Ovis aries) on St Kilda: a consequence of parasite distribution? Oikos 96: Illius AW, Gordon IJ (1990) Constraints on diet selection and foraging behaviour in mammalian herbivores. In: Hughes RN (ed) Behavioural mechanisms of food selection. Springer, Berlin Heidelberg New York, pp Jewell PA, Milner C, Boyd JM (1974) Island suvivors. The ecology of the Soay sheep of St Kilda. Athlone Press Kyriazakis I, Tolkamp BJ, Hutchings MR (1998) Towards a functional explanation for the occurrence of anorexia during parasitic infections. Anim Behav 56: Lafferty KD (1992) Foraging on prey that are modified by parasites. Am Nat 140: Lawes Agricultural Trust (1993) Genstat 5 reference manual. Clarendon, Oxford Lillesand TM, Kiefer RW (1979) Remote sensing and image interpretation. Wiley, New York Lozano GA (1991) Optimal foraging theory: a possible role for parasites. Oikos 60:

8 460 Maizels RM, Holland MJ, Falcone FH, Zand XX, Yazdanbakhsh M (1999) Vaccination against helminth parasites the ultimate challenge for vaccinologists? Immunol Rev 171: Milner JM (1999) Survival, natural selection and foraging efficiency in Soay sheep on St Kilda. PhD thesis. University of London, London Mylrea PJ, Hotson IK (1969) Serum pepsinogen activity and the diagnosis of bovine ostertagiasis. Brit Vet J 125: Newman JA, Parsons AJ, Harvey A (1992) Not all sheep prefer clover: diet selection revisited. J Agric Sci 119: Patterson HD, Thompson R (1971) Recovery of inter-block information when block sizes are unequal. Biometrica 58: Pella E, Colombo B (1973) Study of carbon, hydrogen and nitrogen determination by combustion-gas chromatography. Mikrochim Acta 1973: Penning PD, Rook AJ, Orr RJ (1991) Patterns of ingestive behaviour of sheep contiuously stocked on monocultures of ryegrass and clover. Appl Anim Behav Sci 31: Poppi DP, MacRae JC, Brewer A, Coop RL (1986) Nitrogen transactions in the digestive tract of lambs exposed to the intestinal parasite, Trichostrongylus colubriformis. Brit J Nutr 55: Rook AJ (1998) Design and sampling issues in grazing studies. Proceedings of the IXth European intake workshop, North Wyke. North Wyke Agricultural College, North Wyke, UK Schowengerdt RA (1983) Techniques for image processing and classification in remote Sensing. Academic Press, London Sykes AR (1987) Endoparasites and herbivore nutrition. In: Hacker JB, Ternouth JH (eds) Nutrition of herbivores. Academic Press, Harickvale, pp Zar JH (1984) Biostatistical analysis. Prentice-Hall International, N.J.