1The maintenance or modification of grazing regimes is frequently advocated to deliver conservation targets in pastoral landscapes, but there are few quantitative studies of the effects of grazing on upland birds. This is particularly true with respect to grazing management in agri-environment schemes.
2Numbers of black grouse Tetrao tetrix and their breeding success were therefore monitored at 20 sites in the north of England from 1996 to 2000. Ten treatment sites included areas where grazing was reduced before and during the study to < 1·1 sheep ha −1 in summer and < 0·5 sheep ha −1 in winter. Each was paired with a reference site that held sheep at two (summer) to three times (winter) the density on the experimental sites. The reduced grazing sites ranged from 0·4 to 3·2 km 2 in size and most were part of existing agreements within agri-environment schemes that had been in place for 1–5 years before 1996.
3Numbers of black grouse males displaying increased by an average of 4·6% (SE = 2·1) year −1 at the 10 sites with reduced grazing. Displaying male trends differed significantly between treatment and normally grazed reference sites, where numbers declined annually on average by 1·7% (SE = 1·4).
4Summer black grouse hen densities showed the greatest rate of increase where grazing was restricted on smaller areas of ground (0·4 km 2 ). Declines occurred at sites where the area of restricted grazing exceeded about 1 km 2 . The rates of change in population density, as indicated by numbers of displaying males, peaked in the early years of grazing reduction and then declined after c. 5–7 years.
5The proportion of females that retained broods during the late chick-rearing period was 54% (SE = 0·06) at sites with reduced grazing, significantly greater than the 32% (SE = 0·06) at normally grazed reference sites. There was no difference in the size of broods between grazing treatments.
6This study demonstrates that agri-environment schemes, which encourage extensive management of grazing land, can benefit at least some organisms of conservation importance and lead to some recovery of populations. There is a need, however, for further understanding of how such benefits can be maintained at a landscape scale and over the greater time scales involved in vegetation dynamics and bird population processes.
The maintenance or modification of grazing regimes is frequently advocated to deliver conservation targets within pastoral landscapes (Brown & McDonald 1995; Ward, MacDonald & Matthew 1995). There are, however, few quantitative studies on the relationships between grazing and birds (Fuller 1996). An influence of temporal changes in grazing regime has been demonstrated for the redshank Tringa totanus (L.) on salt marshes, where increased grazing intensity led to reduced breeding densities (Norris et al. 1998). Coarse changes in vegetation type, from heather Calluna vulgaris (L.)-dominated swards to grassland under high levels of grazing, is given as one of the most important factors in the decline of red grouse Lagopus lagopus scoticus (Latham) in Britain (Hudson 1995). For black grouse Tetrao tetrix (L.), correlative studies also in the British uplands suggest reduced population densities and breeding success can be associated with the shorter sward heights maintained by higher densities of grazing animals (Baines 1991, 1996a). Although the stocking rates of grazing animals have varied historically, agricultural intensification in the British uplands, especially during the latter half of the 20th century, has led to the degradation of important moorland and moorland fringe habitats (Thompson et al. 1995). Since 1991, the UK government has funded agri-environment schemes that can offer incentives to reduce densities of grazing livestock. Amongst the principal aims of these schemes has been the reversal of habitat degradation and fragmentation, and the improvement and extension of wildlife habitats (MAFF 2000a).
Formerly widespread in Britain, numbers of black grouse have declined and their range has contracted throughout the 20th century (Baines & Hudson 1995). The most recent of two ornithological atlas surveys found them in 28% fewer 10-km squares in 1988–91 than in 1968–72 (Gibbons, Reid & Chapman 1993). Concern about declining numbers and a contraction in the range occupied by black grouse in Britain has led to several conservation initiatives (Hughes et al. 1998). A similarly unfavourable status is also found throughout much of the west European range (Hagemeijer & Blair 1997; Loneux & Ruwet 1997). Over 95% of black grouse in England are now found within approximately 2500 km2 of the North Pennines, between latitudes 54°10′ N and 55°0′ N, where the population was estimated to include about 1700 displaying males (95% confidence limits 800–3100) in 1995–6 (Hancock et al. 1999). In the North Pennines, most black grouse are found between 200 and 650 m above sea level, where they occur on and close to moorland using a mix of habitats that includes blanket bog, rough grassland, heather moor, herb-rich meadows and pastures (Baines 1994; Starling-Westerberg 2001). One of the principal land uses is the rearing of livestock, notably sheep. Numbers of sheep have increased in the English uplands since the 1970s, associated with agricultural productivity incentives (Anderson & Yalden 1981; Fuller 1996), and they are thought to have contributed to the deterioration of conditions for black grouse (Hughes et al. 1998).
Using agri-environment schemes, a regional conservation initiative for black grouse in the North Pennines has encouraged the reduction of grazing intensity on habitats occupied by black grouse over a wide geographical area (Brown 1999). This paper reports on changes in population density and breeding success in black grouse in response to localized reductions in grazing under such schemes in the uplands of northern England. It compares these changes with concurrently monitored reference sites and more generally considers the effectiveness of such schemes as a conservation tool.
Twenty monitoring sites were selected across the breeding range of black grouse in the north of England (Fig. 1). Each site encompassed between 1·0 km2 and 7·4 km2 (mean 2·6 km2) of moorland and moorland fringe. Ten of these sites included areas where agreements restricted the number of grazing domestic animals, and are referred to as treatment sites. The area covered by restrictive grazing agreements within the treatment sites ranged from 0·4 km2 to 3·2 km2 and the agreements had been effective for between 1 and 5 years before 1996 (Table 1). Each treatment site was paired with a reference site (= control) that was not subject to such restrictions. The mean distance between paired sites was 9·3 km (range 5–21 km). The pairing of sites with contrasting grazing treatment helped ensure that the influence of some extrinsic factors, notably local weather, was comparable for each member within a pair. The selection of treatment sites was restricted by two criteria: the sites had to be occupied by black grouse and had to have recently reduced numbers of grazing animals on at least part of their area. The selection of paired reference sites was a necessary compromise between sufficient proximity to treatment sites to expect comparable weather and yet have sufficient separation distance to expect a high degree of independence between their population trends; a minimum distance of 5 km between all study sites was chosen at the onset of monitoring.
Table 1. Descriptive statistics for areas where grazing had been reduced on black grouse monitoring sites in northern England
Area of reduced grazing (ha)
Perimeter of area with reduced grazing (km)
Year of grazing reduction
Proportion of monitored site subject to reduced grazing (%)
Distance to reference site (km)
The principal grazing animals were sheep Ovis (domestic), predominantly of local hill breeds (Swaledale and blackface). Numbers of domestic grazing animals on study sites were counted eight times between 1998 and 2000 to quantify numerical and seasonal differences between treatments. Small numbers (< 30) of cattle Bos (domestic) grazed in summer on five treatment and three reference sites. For the purposes of assessing stock numbers, one cow was considered the equivalent of 10 sheep, in accordance with agricultural guidelines (MAFF 2000b). There were no wild large herbivores present, other than a very small number of roe deer Capreolus capreolus (L.) that will have made an insignificant contribution to grazing pressure. Most of the monitored sites were within sporting estates (eight treatment and seven reference) where predators were actively controlled. Rough indices of avian predators suggested low densities (medians of 2·7 corvids and 0 raptors km−2 seen during brood counts, see below); as no measurement of mammalian predator abundance was attempted, the influence of predator abundance was not considered further but was assumed to have been similar between grazing treatments.
Point sampling equivalent to the ‘sward stick’ method (Stewart, Bourn & Thomas 2001) was used to assess vegetation height and dominant species composition at each study site between mid-June and mid-July in 1999. Measurements were taken at a minimum of 100 evenly spaced points along transects that traversed each study site. Between one and four parallel transects traversed each site (200–700 m apart) to give representative coverage of all grazing units included. Vegetation height was measured in situ (as opposed to stretching to maximum height) using a vertically held cane. This was the maximum height of the foliage for grasses and herbs, or the whole plant for dwarf shrubs. The most abundant species within a 2-cm radius of the cane was recorded. For analyses, vegetation was classed into one of six groups: ericaceous (overwhelmingly Calluna vulgaris in the north of England); grass (Graminacea); Eriophorum species; Juncus squarrosus (L.); tall Juncus species; ‘other’. The latter included herbs, typically Gallium saxatile (L.), Polytricum moss and unvegetated soil and peat. Compositional analysis of the log-ratios with ‘other’ as the denominator, using a paired manova and the Wilk's lambda statistic, was used to assess the significance of any difference in vegetation composition between grazing treatments (Aitchison 1986). The significance of differences in mean vegetation height and also its coefficient of variation (standard deviation/mean) between treatments was assessed using paired t-tests.
Numbers of displaying male grouse were counted twice at each site within 2 h after dawn between early April and late May, to coincide with the peak and most consistent attendance by males at leks (Baines 1996b). At least 2 weeks separated counts at each site, the highest count for each site being that used to estimate the number of males present. Leks were counted at four treatment and seven reference sites in 1996, and all 20 sites were counted annually from 1997 to 2000.
Indices of female grouse abundance on the study sites were obtained by systematic searching with pointer dogs. Searches were undertaken in spring (mid-March to mid-April) and in summer (mid-July to mid-August), the latter while searching for broods (see below). Spring densities were assessed for eight treatment and five reference sites in 1996 and for all sites from 1997 to 2000.
To assess whether the reduction of grazing levels was associated with changes in the abundance of black grouse, the rates of change in three measured indices of population density were compared between treatment and reference sites. Trends for each site were determined through generalized linear modelling, regressing the annual counts against year and assuming a Poisson error distribution. As there was a maximum of only five counts per site, for simplicity a linear trend was assumed. A paired t-test compared the trend coefficients between grazing treatments. The trend coefficients gave the instantaneous rates of change, from which annual rates were calculated by exponentiating.
Variation in the areas covered by, and dates of entry into, the restrictive grazing agreements within the treatment sites gave an opportunity to look for any influence of size and age of such lightly grazed swards. The differences in the trend coefficients between pairs of sites (coefficient for the treatment site less that for the reference) were plotted against the area of land within a restrictive grazing agreement at the treatment sites, and also the number of years for which grazing had been restricted (as of 1998, to give the mean associated with the 5-year 1996–2000 trends). The between-pair differences provided a measure of the deviation from the reference situation, allowing any spatial or temporal influence of grazing reduction to be assessed. As well as the absolute area of land entered into a restrictive grazing agreement, the ‘shape’ of the plot may also exert an important influence. Accordingly, the differences in trend coefficients were also examined against a measure of the ‘grain’ of agreement areas [length of edge (km)/area (km2)].
Breeding success was estimated by systematically searching monitored sites with pointer dogs for female black grouse and noting brood sizes. Such searches covered each entire site and were undertaken between mid-July and mid-August, when most young were expected to be more than 3 weeks old (after the main period of pre-fledging mortality) and before the break-up of identifiable broods (Baines, Wilson & Beeley 1996). Breeding success was measured for three treatment sites and five reference sites in 1996 and all sites were monitored annually from 1997 to 2000. The number of females encountered on a site visit ranged from 0 to 13 (median 3). Trends in brood density were compared between grazing treatments as for the three indices of adult density (see above).
To assess whether reduced grazing was associated with increased breeding success, the proportion of females found attending broods and the brood sizes were compared. Three ratios (broods per female seen, chicks per brood and chicks per female seen) were log-transformed and weighted by the number of females, or broods as appropriate, prior to comparison. The means across all years of the three measures for each site were compared using paired t-tests (treatment vs. reference). The nature of any relationships with the extent of a restrictive grazing agreement, its grain and the length of time for which grazing had been restricted, was investigated by plotting the between-pair differences in the ratios against area, or age, as with the population trends (see above). Instances when no females were counted were treated as missing data, as in no case were chicks found at sites where no females were found.
GRAZING AND VEGETATION
Maximum summer (April–September) grazing levels on agreement areas within treatment sites were the equivalent of 1·1 sheep ha−1 compared with 2·4 sheep ha−1 on reference sites [means = 0·5 (SE = 0·1) and 1·7 (SE = 0·2), respectively, paired t9 = 7·85, P < 0·001 (treatment vs. reference)]. In winter, the maximum density on agreement areas within treatment sites was the equivalent of 0·5 sheep ha−1, with 1·6 on reference sites [means = 0·2 (SE = 0·1) and 1·0 (SE = 0·1), respectively, paired t9 = 6·27, P < 0·001].
The composition of vegetation did not differ between grazing treatments (paired manova of log-ratios, Λ= 0·67, P= 0·28), with grasses averaging 45% cover and the other groups less than 20% each (Fig. 2). Vegetation height was greater at treatment sites, within the areas subject to grazing restrictions (mean height = 33·6 cm, SE = 1·6), than at reference sites (21·5 cm, SE = 1·6) (paired t9 = 6·0, P < 0·001). There was no difference in vegetation height between areas within treatment sites, where grazing remained unrestricted (mean height = 22·0 cm, SE = 2·1), and reference sites. The restrictive grazing agreements were the only areas where the mean sward height in summer exceeded 30 cm (as measured using the sward stick method). There was a trend for greater heterogeneity in vegetation height at reference sites (mean coefficient of variation = 0·78, SE = 0·05) than within areas of reduced grazing at treatment sites (mean coefficient of variation = 0·66, SE = 0·06); however, the difference was not significant (paired t9 = 1·95, P= 0·08).
TRENDS IN BIRD NUMBERS
Of the three indices of population density for black grouse, comparison of trends showed a significant difference between treatments only for the number of displaying males in spring (Table 2). At sites that included areas of restricted grazing (treatment sites), displaying males increased by an average of 4·6% (SE = 2·1) year−1 compared with a concurrent average annual decline of 1·7% (SE = 1·4) at reference sites (Fig. 3). In 1997, the first year that all sites were counted, a total of 110 displaying males was counted on treatment sites (mean = 11·0, SE = 3·6). At reference sites the comparable figure was 140 (mean = 14·0, SE = 1·9). In 2000, the mean number of displaying males at treatment sites was 14·0 (SE = 4·6) and at reference sites 11·1 (SE = 1·8).
Table 2. Comparison of trends in three indices of black grouse population density and brood density between grazing treatments in northern England, 1996–2000. Treatment sites included areas where grazing had been reduced. Reference sites retained relatively high levels of grazing throughout
The trend coefficients are derived from regression of annual counts by year for each site and give the instantaneous rate of change. The means and standard errors are those of the 10 coefficients (from 10 sites) within each grazing treatment.
Displaying males at leks
1·52 (± 0·74)
−0·50 (± 0·36)
t9 = 2·61, P = 0·03
Females in spring
0·67 (± 0·65)
−0·46 (± 0·65)
t9 = 1·56, P = 0·16
Females in summer
−0·03 (± 0·30)
−0·50 (± 0·25)
t9 = 1·17, P = 0·27
0·10 (± 0·07)
0·12 (± 0·19)
t9 = 0·08, P = 0·94
The mean encounter rate of females during the spring across all 5 years at treatment sites was 2·7 km−2 (SE = 0·6), while at reference sites it was 2·1 km−2 (SE = 0·3) (paired t-test of log-transformed mean densities, t9 = 0·65, P = 0·53). During the summer, the mean encounter rate of females at treatment sites was 2·2 km−2 (SE = 0·5), while at reference sites it was 1·3 km−2 (SE = 0·2) (paired t-test of log-transformed mean densities, t9 = 0·83, P = 0·43). Although there were no significant differences in trends between treatments for either index of female density, their relative difference was the same as for displaying males, a suggested increase at treatment sites compared with that for reference sites (Fig. 3).
Plotting the between-pair difference in trend coefficients of female abundance in summer against area within a restrictive grazing agreement (effectively the area where average summer vegetation heights exceeded 30 cm), or its grain, suggested the effect of treatment was least where that treatment was greatest in extent [linear regression of between-pair differences in trend coefficients with area (log-transformed), F1,8 = 6·75, P= 0·03, R2 = 0·47) or where the perimeter to area ratio was least (F1,8 = 5·67, P= 0·04, R2 = 0·42)] (Fig. 4). Indeed, the deviation from the references barely exceeded zero where the restrictive grazing agreement areas exceeded about 100 ha in extent or the perimeter to area ratio was less than about 4 (Fig. 4). No relationships with agreement area were apparent for numbers of displaying males or the numbers of females observed in spring.
A quadratic relationship was suggested, although marginally non-significant (F1,7 = 3·67, P = 0·08, R2 = 0·51), for trends in displaying males (the most reliable of the three population indices investigated) with the length of time since entry into a restrictive grazing agreement (Fig. 5). The rate of increase in the index peaked at about 5–7 years and declined thereafter.
Fifty-four per cent of the females encountered at treatment sites during the summer were with broods. On the reference sites, the comparable proportion was significantly lower at 32% (Table 3). There was no difference in brood size between the two treatments, however (Table 3). Although breeding success varied between years, brood survival was consistently greater at treatment sites, although the difference was not statistically significant for every year (Fig. 6). There was no apparent relationship between breeding success and either the area, grain or age of restrictive grazing agreements.
Table 3. Comparison of parameters of black grouse breeding success between grazing treatments in northern England, 1996–2000. Treatment sites included areas where grazing had been reduced. Reference sites retained relatively high levels of grazing throughout
Tests were performed on the ratios after log-transformation.
Proportion of females with broods
0·54 (± 0·06)
0·32 (± 0·06)
t9 = 2·61, P = 0·03
Number of chicks per brood
2·70 (± 0·11)
2·60 (± 0·21)
t9 = 0·78, P = 0·47
Number of chicks per female
1·41 (± 0·25)
0·82 (± 0·21)
t9 = 3·12, P = 0·02
Grazing reductions within sites in northern England were associated with more successful breeding by, and increasing population densities of, black grouse (as indicated by the proportion of females rearing broods and numbers of displaying males) relative to sites where grazing remained relatively high throughout. Changes in the indices of female population density in spring and summer showed a similar, although non-significant, tendency. The latter were derived by sampling fewer individual birds (mean of 98 in spring and 66 in summer for all 20 sites combined per year for the 4 years 1997–2000) compared with the numbers of displaying males sampled (mean of 245). The trends determined for female indices are therefore likely to be less robust than those for displaying males.
The grazing trials monitored in this study relied on established financial incentives (agri-environment schemes) that defined maximum stocking levels at those sites where grazing was restricted (MAFF 2000b) or, alternatively, were private agreements with comparable restrictions. The management of the remaining (reference) sites was determined by the general carrying capacity for stock and other broadly implemented agricultural subsidies based on headage payments (Brouwer & Van Berum 1996). The latter are limited by the operation of quotas and thus, along with the more restricted regimes of the agri-environment schemes, effectively established pairs of sites with bimodal grazing treatments. In this study, no attempt has been made to determine any stocking densities that may prove ‘optimal’ for black grouse. However, the observed differences between treatments do demonstrate that degradation of habitats for black grouse by intensive grazing can be reversible and that manipulation of grazing regimes can contribute towards their conservation.
Taller vegetation can provide better feeding and chick-rearing opportunities for black grouse, associated both with increased vegetative food availability and increased invertebrate abundance (Baines 1996a). In this study, the actual mechanisms leading to the observed differences in breeding success are not known. As there was no detectable difference in the composition of dominant vegetation types between grazing treatments, the responses by black grouse were most likely related to the resultant differences in vegetation structure. In addition to the measured height difference, other factors contributing to the beneficial effects might include the increased abundance of seeds and other fruiting bodies with associated increases in phytophagous invertebrates (Wakeham-Dawson & Smith 2000), and possibly also the increased abundance of scarce plants [for example bilberry Vaccinium myrtilus (L.)] that would remain undersampled by the method used to survey vegetation but would still be available as food for black grouse.
Weather (Moss 1986; Loneux, Lindsay & Ruwet 1997) and predation (Kahaula, Helle & Helle 2000) have both been implicated as influencing black grouse abundance. As well as affecting food availability, vegetation structure could exert an influence with these other factors through interaction (Thirgood et al. 2002). Tall vegetation may provide shelter from adverse weather and cover against avian predators. Conversely, taller vegetation could exert a negative influence by reducing opportunities for chicks to dry out during wet weather (increasing the risk of hypothermia) and also potentially increasing the risk from ground predators that rely on olfactory detection through enhanced stalking opportunities afforded by thicker cover. Some of the thickest vegetation might also have a reduced associated abundance of invertebrates because of shade-induced reduced temperatures close to ground level (Wakeham-Dawson & Smith 2000). The prediction of changes in bird populations associated with habitat changes requires knowledge of the mechanisms that determine their distribution and use of the patches that contribute to a mosaic of habitats (Bernstein, Krebs & Kacelnik 1991). Our study set out to determine whether simple grazing reductions could influence black grouse demography and therefore be used as an effective conservation tool. The reduced response in female density during the late brood-rearing period with increased area of tall vegetation, but the lack of an apparent relationship with breeding success suggests, a tendency to avoid extensive tall swards for breeding, or to leave such areas following breeding failure. Although no similar trend for brood densities was apparent, this may well be a result of the small sample size during the study period (mean of 1·6 broods site−1). The key periods for breeding failure, whether nest or chick losses, were expected to be prior to the sampling of broods in this study (Baines, Wilson & Beeley 1996). Although reduced grazing appears to have benefited black grouse, at least during the time scale of this study, there is a need to further elucidate the mechanisms through which the effect operates and, critically, to determine if its extent can become limiting. The home range of broods in the North Pennines can be small (under 50 ha; Starling 1990). It follows that the establishment of all requirements for chick rearing may need to be on a similarly fine scale. Within our study sites in the north of England, the patches of relatively tall vegetation (those within restrictive grazing agreements) approached sizes where there appeared to be a redistribution of some individuals, that is females avoiding extensive areas in summer. However, from the observed changes in numbers of displaying males, there was, as yet, no suggestion of an influence on local population density. In reality, the grain of vegetation structure is likely to be important, that is a measure of the distance between contrasting sward heights and the ability of chicks to travel between them. In this study there was a strong correlation between perimeter length and area of tall swards and so their relative influences were confounded.
Our suggested temporal change in the influence of grazing restriction clearly requires further investigation. Although the differences in rates of population change were least for the older agreements relative to their reference, this may actually be a density-dependent response by the birds as the carrying capacity of these areas is approached. Longer term monitoring could identify whether conditions for the birds ultimately deteriorate. Such a quadratic response in populations is often reported following grazing exclusion and the planting of trees on moorland (Cayford 1993). This is often attributed, on no firm basis, to canopy closure but may well involve other factors, such as adverse changes in ground vegetation cover.
The response of vegetation to reductions in grazing can be variable and depends on the degree and time scale of its degradation and also the type of grazing animals involved (Hill, Evans & Bell 1992). All our monitored sites supported black grouse at the start of the study, and so it can be assumed that any existing degradation had been neither too severe nor too long-term as these sites still retained sufficient attributes to hold the species. The initial and principal aim of the grazing reductions on the treatment sites was to increase the extent of heather cover, and they were typically targeted at land where heather already occurred but comprised less than 25% of the ground cover at the onset of stock reductions (MAFF 2000b). Although there was no evidence that this principal aim was being achieved in all cases, all sites where grazing had been reduced included at least some ericaceous vegetation. All ground would not be expected to respond similarly, either in terms of vegetation cover or associated avifauna. Areas that have become unsuitable for black grouse and from which the species has been absent for a number of years may actually be more severely degraded and may not necessarily respond to simple grazing reductions in a similar manner.
The role of agri-environment schemes in encouraging extensive agriculture is thought to have benefited the ‘near-threatened’ little bustard Tetrax tetrax L. in southern France (Wolff et al. 2001). That study acknowledges the importance of habitat diversity and an appropriate degree of landscape fragmentation for that species. Our monitoring of black grouse shows that a species of immediate conservation concern can benefit from agri-environment schemes that encourage extensive management of grazing land. However, as well as ensuring that an appropriate diversity of habitats is maintained within a home range occupied by a population of black grouse, management should also be on a sufficiently fine scale to maintain a diversity of vegetation structures to encourage successful breeding. Within the English uplands, the extent of land being entered into restrictive grazing agreements through agri-environment schemes is increasing, and with greater funds becoming available (MAFF 2000a) this is likely to continue for at least a number of years. The consideration of the implementation of such management has therefore become pertinent. Widespread and contiguous coverage with similar grazing management, even if equivalent to the reduced levels on the treatment sites in this study, will not necessarily prove to be best practice for black grouse or indeed other sympatric species. There is a need for further understanding of how agri-environment and related conservation policies can maintain species and species diversity on a landscape scale and also over the greater time scales in which vegetation dynamics and, ultimately, the population processes of animals take place.
This study was part of the monitoring programme of the North Pennines Black Grouse Recovery Project supported throughout by English Nature, the Game Conservancy Trust, the Ministry of Defence and the Royal Society for the Protection of Birds. Supplementary support from National Wind Power is gratefully acknowledged. Jackie Longrigg, Jonathan Pratt and James Philips undertook some of the fieldwork. Chris Wernham of the BTO facilitated some of the analyses. Nicholas Aebischer advised on statistics and, along with Robert Manners, two referees and the editor, commented on earlier drafts. We appreciate the support and co-operation of the many land owners and managers on whose ground we worked.