Effects of pheasant management on vegetation and birds in lowland woodlands


Correspondence author. E-mail: rdraycott@gct.org.uk


  • 1Releasing pheasants in woodlands for game shooting is a widespread practice in the British countryside. Apart from changes to ground flora in woodland release pens, the effects of pheasant releasing on woodland biodiversity are poorly understood.
  • 2We surveyed 159 lowland broad-leaved woods in southern and eastern England during spring–summer 2004 to determine the impact of pheasant management on vegetation structure and composition and on bird populations. Eighty-one of the woods contained pheasant release pens with annual releases of pheasants and associated gamebird management such as supplementary feeding in winter. In the remaining 78 woods there had been no pheasant management for at least 25 years.
  • 3We found that the pheasant-managed woods had a more open structure, with between 2% and 7% less canopy cover and a denser field layer with between 5% and 58% more ground vegetation cover, including grasses and bramble Rubus fruticosus.
  • 4We recorded approximately 40% more birds in woods in southern England and between 22% and 32% more birds were observed in pheasant-managed woods than control woods. Woodpigeons and warblers were more abundant in pheasant-managed woods.
  • 5Synthesis and applications: We conclude that the impacts of pheasant releasing on vegetation structure and bird communities in woodlands are benign or positive. This study demonstrates that some aspects of woodland management for pheasants, including reducing the extent of canopy cover can encourage growth of understorey vegetation which helps create favourable conditions for some woodland bird species.


There is a long tradition of hunting game birds and mammals within Europe. This continues today, taking many forms, and can have major economic and social impacts on rural areas (Lecocq 2004). The motivators for game harvesting are diverse and include subsistence hunting, population management, diversification of farm income and recreation. Where economic and recreational drivers are involved, habitat and predator management are often implemented to increase populations of game. With increased agricultural intensification in recent decades and the relative ease of rearing game birds artificially, releasing of birds for shooting is now widespread in many European countries (Draycott, Pock & Carroll 2002). Despite the prevalence of game hunting, the costs and benefits to other wildlife sharing the same habitats are poorly understood or have been documented only at a local scale.

In Britain, common pheasants Phasianus colchicus L. have been associated closely with woodlands since the 16th century, when they became important as a quarry species (Longrigg 1977; Hill & Robertson 1988). In the 1960s the development of artificial rearing methods led to a substantial increase in the number of pheasants released (Tapper 1992). This increase continued until the early 1990s, and at least 25 million pheasants are currently released in British woods each summer (Tapper 1999). There is concern among conservationists about the effects of this widespread practice on native wildlife (Sutherland et al. 2006), but there is little published research concerning this issue (Fuller et al. 2005).

Pheasants are released typically via woodland release pens during July–August, at 6–8 weeks of age, from which they gradually disperse as they become accustomed to their environment (Sage, Ludolf & Robertson 2005). These areas are where negative ecological consequences of pheasant releasing are likely to be most obvious. For example, Sage, Ludolf & Robertson (2005) recorded detrimental effects to sensitive ground flora within pheasant release pens in ancient seminatural woodlands when they contained high densities of pheasants (> 1500 birds per ha of pen). However, effects within the remainder of the wood are not well understood.

Despite a national trend of loss of large areas of farm woodland to agriculture and urbanization in the second half of the 20th century (Fuller 1995; Peterken 1996) and a reduction in the levels of management in remaining farm woodlands because they were often no longer economically viable (Fuller et al. 2005), game managers have retained, managed and planted woodlands for the specific purpose of providing habitat and favourable conditions for pheasants and pheasant shooting (Cobham Resource Consultants 1997; Howard & Carroll 2001; Tapper 2005).

Although game managers have long recognized the importance of woods for holding pheasants on their properties through the autumn and winter (Walsingham & Payne-Gallwey 1895), the detailed habitat requirements of pheasants were not determined until relatively recently. Robertson et al. (1993a, 1993b) showed that pheasants preferred small woodlands with a high proportion of shrubby cover 30–200 cm tall and a shrubby woodland edge. Depending on the type of wood, there are a number of ways in which these conditions can be created or enhanced. These include traditional coppicing, sky-lighting (localized reduction in extent of canopy cover) to create favourable conditions for the regrowth of the understorey vegetation or planting shade-tolerant shrubs in woodlands with closed canopies (Ludolf, Robertson & Woodburn 1989). These are techniques which are likely to benefit other wildlife, including songbirds (Fuller 1995) and butterflies (Robertson, Woodburn & Hill 1988; Ludolf, Robertson & Woodburn 1989). Many woodland songbirds, particularly migrant species, occur in greater densities in shrubby woodland (Fuller & Henderson 1992). Recent research has shown that many of these woodland species are declining, and a reduction in the amount of understorey vegetation in woodlands has been cited as a possible cause of these declines (Fuller et al. 2005; Amar et al. 2006). While it is generally recognized that managing woodlands for pheasants may therefore benefit aspects of woodland biodiversity, there is little direct evidence for or against this (Fuller et al. 2005).

In this study we quantify the effects of pheasant releasing and its associated management on the vegetation structure and composition, and on the bird community, of the interior of lowland broad-leaved woodland by comparing woods with and without pheasant releasing. We investigate relationships between songbird numbers and woodland parameters in relation to pheasant management.

Materials and methods

study woods

We surveyed 159 woods in southern and eastern England during 11 May–15 July 2004. We restricted the geographical range of our study sites to the Hampshire and South Wessex Downs Natural Areas in the south (51°10′ N, 1°41′ W) and the East Anglian Plain, Breckland and the Suffolk Coast and Heaths Natural Areas in East Anglia (52°15′ N, 0°15′ E). Eighty-one of the woods were managed for pheasant shooting. Regions were surveyed simultaneously. Our criteria for defining woods as managed for pheasants were that there was an annual release of pheasants in the woods and supplementary food was provided for them in winter. In the remaining 78 ‘control’ woods, no pheasant releasing or supplementary feeding had been undertaken for at least 25 years. The minimum size of woods was 6·5 ha and pheasant and control woods were split evenly between regions. To select woods we used a geographic information system (GIS) (MapInfo 7·0; Mapinfo Corporation 2002) which selected randomly 90 1-km2 plots in each region that contained at least 10 ha of deciduous woodland. Some of these woods were discounted because individual blocks of woodland within the square were too small or were just shelterbelts, while others were discounted because access permission could not be obtained. This left 31 randomly selected pheasant woods and 33 randomly selected control woods which were surveyed. These comprised 40% of the pheasant woods and 36% of the control woods in the southern region and 38% and 48%, respectively, in the eastern region. We identified the remaining woods from a variety of sources and included woods owned or managed by Forestry Commission, National Trust, GCT members, County Councils and private landowners. Although these woods were not selected randomly, we had no prior knowledge of their vegetation characteristics or bird populations. A maximum of two pheasant and two control woods were surveyed on the same site. In 40% of cases a single wood per site was surveyed. All woods fell within five National Vegetation Classification (NVC) types (Hall et al. 2004), the majority (70%) being either W8 (ash Fraxinus excelsior L., field maple Acer campestre L. and dog's mercury Mercurialis perennis L.) or W10 (oak, Quercus robur L., bracken Pteridium aquilinum L. and bramble Rubus fruticosus L.).

survey methods

We worked in a 4-ha plot within each wood. A team of two surveyors visited each wood together. One surveyor collected vegetation data and recorded signs of deer activity while the other concentrated on bird counts. We located each plot simply by entering the wood at the first access point we came to and then heading 20 m away from the access point before commencing north–south transects. We avoided the woodland edge as this is a distinct habitat (Fuller 1995) and restricted our survey to the woodland interior (> 20 m from the edge). We measured vegetation at a total of 40 points spaced 20 m apart along four equally spaced transects within the 4-ha plot. At each sampling point we estimated ground cover to the nearest 5% and recorded presence or absence of bare ground, dead wood, moss, herbs, grass, shrubs, tree seedlings and trees within a 1-m2 quadrat. To obtain a measure of vegetation structure we noted the presence of vegetation in a column 30 cm in diameter at six height bands (0–10 cm, 11–30 cm, 31 cm−1 m, 1–2 m, 2–5 m and above 5 m).

During this time we recorded the numbers of adults of each bird species seen or heard in the 4-ha plot. Care was taken to avoid repeat counts of the same individuals. Any juveniles were excluded owing to the difficulty of obtaining a complete count of each brood. All surveys were conducted between 0800 h and 1600 h. To obtain estimates of deer abundance and activity, we recorded the number and species of deer observed in the study plots. We also recorded and identified the droppings and tracks of deer and signs of deer browsing, bark-stripping and rubbing at sampling stations and within a 1-m wide transect between sampling stations. We recorded a maximum of 40 records of each sign of deer activity per study plot (i.e. one per sampling station).

statistical analysis

Our measures of ground cover and the presence of different vegetation types at ground level and at each height category from the 40 quadrats were combined to give a mean proportion for each variable per study plot. The use of 40 quadrats in each plot improved the estimate of each of our measured variables by accounting for the nonrandom distribution of plants (Sage, Ludolf & Robertson 2005). These proportions were transformed to angles (arcsine inline image) prior to analysis. Because numbers of individual bird species were low, analyses were performed on the following groups: total birds recorded, total birds excluding woodpigeons Columba palumbus L., tits, finches, warblers, ground feeders, woodpigeons and others (see Table 1 for a full list of the species within each group). The bird data were analysed as log10(x + 1)-transformed numbers per site. We also analysed species richness (number of different species observed at each site).

Table 1.  Mean ± SE numbers of birds recorded in 4 ha woodland plots in pheasant and control woods in southern and eastern England during May–July 2004
 Southern woodsEastern woodsWald statistics (P-values) for effects of region, treatment and interaction
Pheasant woods
(n = 41)
Control woods
(n = 38)
Pheasant woods
(n = 40)
Control woods
(n = 40)
(1 d.f.)
(1 d.f.)
Region × treatment
  • a

    Analyses restricted to randomly selected woods (southern region: pheasant woods n = 16, control woods n = 14, eastern region: pheasant woods n = 15, control woods n = 19) (gamebirds excluded from total). Tits = blue tit Parus caeruleus L., great tit P. major L., marsh tit P. palustrs L., long-tailed tit Aegithalos caudatus L., coal tit P. ater L. Finches = chaffinch Fringilla coelebs L., greenfinch Carduelis chloris L., Bullfinch Pyrrhula pyrrhula L. Warblers = blackcap Sylvia atricapilla L., willow warbler Phylloscopus trochilus L., chiffchaff Phylloscopus collybita Vieillot, garden warbler Sylvia borin Boddaert, whitethroat Sylvia communis Lathan, goldcrest Regulus regulus L. Ground-feeders = blackbird Turdus merula L., song thrush Turdus philomelos Brehm, mistle thrush Turdus viscivorus L., robin Erithacus rubecula L., dunnock Prunella modularis L., wren Troglodytes troglodytes Swainson. Others = great spotted woodpecker Dendrocopos major L., green woodpecker Picus viridis L., nuthatch Sitta europea L., treecreeper Certhia familiaris L., spotted flycatcher Muscicapa striata Pallas. Pigeon = woodpigeon Columba palumbus L.

Tits 4·6 ± 0·4 4·0 ± 0·4 2·4 ± 0·22·7 ± 0·319·37 (< 0·001)0·31 (0·58) 
Finches 1·4 ± 0·2 1·3 ± 0·2 1·2 ± 0·11·0 ± 0·11·69 (0·19)0·72 (0·396) 
Warblers 1·8 ± 0·3 1·3 ± 0·3 2·2 ± 0·31·3 ± 0·20·70 (0·70)7·68 (0·006) 
Ground-feeders 4·8 ± 0·3 4·5 ± 0·4 3·2 ± 0·32·7 ± 0·220·96 (< 0·001)1·27 (0·26) 
Others 1·0 ± 0·1 0·4 ± 0·1 0·2 ± 0·10·3 ± 0·117·86 (< 0·001)4·03 (0·045)12·93 (< 0·001)
Woodpigeonsa 3·6 ± 0·7 1·6 ± 0·5 2·1 ± 0·60·5 ± 0·220·57 (< 0·001)14·52 (< 0·001) 
Total16·7 ± 0·912·7 ± 1·011·0 ± 0·89·0 ± 0·535·03 (< 0·001)9·28 (0·002) 
Total (excluding pigeons)13·6 ± 0·711·5 ± 0·9 9·0 ± 0·78·1 ± 0·531·8 (< 0·001)4·12 (0·042) 

Vegetation and bird data were analysed using restricted maximum likelihood (REML) meta-analyses. Error variances for sites where one wood was surveyed and sites where more than one wood was surveyed were estimated separately. The REML analysis requires a normally distributed dependent variable, hence the transformations of the vegetation and bird data. Various dependent variables were tested using a full model that included the following factors: region, selection method (random or non-random), treatment (pheasant wood or control wood) and date category (representing three time-periods, each containing approximately the same number of woods 11 May–6 June, 7–30 June and 1–15 July). First- and second-order interactions between region, selection and treatment were included. Site was entered as a random term and the experimental factor was the classification of sites as single wood surveyed or multiple woods surveyed. Wald tests were used to test hypotheses. In cases where selection was significant, we repeated the analysis on just the randomly selected sites to test the validity of the results for other significant factors. For the bird analysis, the following additional factors were included in the model: survey start category (survey initiated before or after 12·00), cloud cover (0–33%, 34–66%, 67–100%), wind (no wind, slight breeze or breezy) and rain (dry or drizzle). We also included a measure of woodland area in the analyses of bird numbers and species richness, as woodland area can influence numbers and diversity of birds (Vanhinsbergh et al. 2002). We calculated the area of woodland within a 1-km radius of the centre of each study plot using the Countryside Information Service data set (Fuller et al. 2000) within MapInfo 7·0. In the analyses of ground cover and bird numbers, several dependent variables were tested using the same model raising the potential problem of multiple testing. Therefore, in these analyses results were considered significant only if P < 0·01. For all other tests, results were considered significant if P < 0·05. To determine the influence of region and treatment on bird assemblages, compositional analysis was undertaken on bird groups following the methods outlined in Aebischer, Robertson & Kenward (1993), but modified to take into account the recommendations of Bingham & Brennan (2004).

Two variables were derived from the data on deer to reflect deer abundance and deer browsing within the woods. A principal component analysis (PCA) was conducted on the counts of deer seen, droppings and footprints and the first principal component score was used in analyses as the measure of deer abundance. The measure of deer browsing was the proportion of survey points at which browsing was recorded. The area of woodland within a 3-km radius of the centre of the survey plot was included in the analyses of deer abundance and deer browsing as available woodland area is likely to influence deer abundance and levels of browsing (Sage et al. 2004). Where treatment was found to significantly influence bird numbers, REML analyses were conducted to explore relationships between bird numbers and vegetation variables to determine whether differences in vegetation composition and structure between pheasant-managed and control woods influenced bird numbers. The full model was as follows: region + selection + region × selection + site + date category + start category + cloud + wind + rain + vegetation cover + bramble cover + canopy cover + deer PCA. Statistics were calculated in genstat 8·2 (Lawes Agricultural Trust 2005).


site and survey characteristics

The samples of pheasant woods and control woods contained similar numbers of each woodland type (combined regions: χ24 = 6·53, P = 0·18, southern region: χ24 = 4·80, P = 0·31, eastern region: χ24 = 4·90, P = 0·30). There was no difference in mean survey dates between treatments (F1,155 = 0·06, P = 0·82) but mean survey date was 5 days earlier in eastern woods (F1,155 = 2·75, P = 0·001). On average surveys took 67 min to complete and there was no difference in duration between treatments, although surveys typically lasted 10 min longer in the southern region (treatment F1,155 = 1·29, P = 0·26, region F1,155 = 17·88, P < 0·001).

ground cover

In the southern region pheasant woods had 6% more ground vegetation cover than control woods and in the eastern region pheasant woods had 58% more ground vegetation cover (Table 2). There were several regional differences in ground cover composition (Table 2). Grasses occurred 29% more frequently in eastern woods and there was a significant region × treatment interaction with respect to forbs; with forbs recorded in 85% more quadrats in pheasant woods in the eastern region while similar levels were recorded in pheasant and control woods in the southern region. Dead wood was recorded in 65% more quadrats in woods in the southern region, while bare ground was recorded in 66% more quadrats in eastern woods. There was also a significant effect of sampling date on frequency of occurrence of dead wood in quadrats, with more observed earlier in the sampling period (Wald = 12·27, d.f. = 2, P = 0·002). Tree density was 2·5 times higher in eastern woods (Table 2). There was a significant region × treatment interaction for tree seedlings with similar frequencies of tree seedlings in pheasant and control woods in the eastern region, but 2·5 times more tree seedlings in control woods than pheasant woods in the southern region.

Table 2.  Frequency of occurrence in quadrats of vegetation type and mean percentage vegetation cover in quadrats in pheasant-managed and control woods in southern and eastern England during May–July 2004
 Southern woodsEastern woodsWald statistics (P-values) for effects of region, treatment and interaction
Pheasant woods
(n = 41)
Control woods
(n = 38)
Pheasant woods
(n = 40)
Control woods
(n = 40)
(1 d.f.)
(1 d.f.)
Region × treatment
  • a

    Analyses restricted to randomly selected woods (southern region: pheasant woods n = 16, control woods n = 14, eastern region: pheasant woods n = 15, control woods n = 19).

Forbsa76·1 ± 6·070·5 ± 5·783·3 ± 3·345·0 ± 6·14·50 (0·034)14·53(< 0·001)8·53(0·003)
Grass23·9 ± 3·120·7 ± 3·738·9 ± 3·930·2 ± 4·210·61 (0·001) 4·02 (0·045) 
Shrub47·7 ± 3·952·0 ± 4·448·9 ± 4·445·5 ± 4·40·62 (0·43) 0·14 (0·71) 
Mossa28·3 ± 1·443·4 ± 5·131·5 ± 5·934·5 ± 6·10·1 (0·76) 3·90 (0·048) 
Dead wood32·8 ± 1·937·9 ± 2·321·3 ± 2·519·7 ± 2·253·22 (< 0·001) 0·35 (0·55) 
Bare ground25·6 ± 2·623·3 ± 3·346·8 ± 4·834·6 ± 4·610·7 (0·001) 2·70 (0·10) 
Tree seedling10·7 ± 2·126·3 ± 3·517·7 ± 2·420·8 ± 2·70·31 (0·58) 9·75 (0·002)7·01 (0·008)
Tree5·27 ± 0·8 5·6 ± 0·812·8 ± 1·414·9 ± 1·555·28 (< 0·001) 1·30 (0·23) 
Mean percentage covera58·0 ± 3·954·5 ± 5·666·9 ± 4·042·1 ± 4·50·47 (0·49)13·79 (< 0·001) 

vegetation structure

In the height category 31 cm–1 m there were significant treatment (Wald = 14·88, d.f. = 1, P < 0·001) and regional (Wald = 21·78, d.f. = 1, P < 0·001) effects. In pheasant woods in the southern region, herbaceous vegetation was recorded in this height category in 64·4 ± 2·8% (mean ± 1 SE) of sampling stations compared with 52·6 ± 3·5% in control woods. In this height category in the eastern region, herbaceous vegetation was recorded in 45·9 ± 4·0% of sampling stations in pheasant woods compared with 37·4 ± 3·9% in control woods. There were no differences in structure of herbaceous vegetation cover between treatment or region in any other height categories.

In the height category 0–10 cm, R. fruticosus was more abundant in pheasant woods (southern region, pheasant woods: 12·8 ± 2·2%, control woods: 7·9 ± 1·7, eastern region, pheasant woods: 18·8 ± 3·1%, control woods: 14·5 ± 2·6%, Wald = 6·00, d.f. = 1, P = 0·01). There was also a significant regional effect with more R. fruticosus recorded in eastern woods (Wald = 4·49, d.f. = 1, P = 0·03). R. fruticosus was also more abundant in pheasant woods in the height category 11–30 cm (southern region, pheasant woods: 26·9 ± 3·7%, control woods: 20·8 ± 3·6%, eastern region, pheasant woods: 21·9 ± 3·3%, control woods: 16·8 ± 2·9%, Wald = 4·35, d.f. = 1, P = 0·04). There were no differences in the abundance of R. fruticosus over 30 cm in height between regions or treatments. There were no differences in abundance of woody shrubs between treatments or regions except in the height category 1–2 m, where there was a significant region × treatment interaction (Wald = 7·74, d.f. = 1, P = 0·005). This was due to more shrubs being recorded in this height category in pheasant woods in the east (pheasant woods: 16·1 ± 1·9%, control woods: 10·9 ± 1·7%) and in control woods in the south (pheasant woods: 19·8 ± 2·0, control woods: 25·0 ± 2·8).

Higher levels of tree canopy cover were recorded in control woods than pheasant woods (southern region pheasant woods: 76·3 ± 1·8%, control woods: 83·6 ± 2·1%, eastern region pheasant woods: 78·2 ± 2·5%, control woods: 80·3 ± 2·4%, Wald = 6·26, d.f. = 1, P = 0·01), indicating that pheasant woods had a more open structure. There was no regional difference in levels of canopy cover (Wald = 0·01, d.f. = 1, P = 0·91), but there was a significant date effect with higher levels of canopy cover later in the survey period (Wald = 17·39, d.f. = 2, P < 0·001).

bird abundance

Overall, between 22% and 32% more birds were observed in pheasant woods than control woods (Table 1). Higher numbers of birds in the ‘warbler’ group, and woodpigeons were recorded in pheasant-managed woods in both regions. There were also regional differences in numbers of many bird groups with a tendency for more birds in southern woods (Table 1). There was a significant region × treatment effect for the ‘others’ group, with higher numbers recorded in pheasant woods in the south and control woods in the east. There was a decline in the total number of birds recorded per wood through the survey period (Wald = 35·11, d.f. = 2, P < 0·001). Date was a significant factor for the finch group with more birds recorded in earlier surveys (Wald = 43·64, d.f. = 2, P < 0·001). There was a significant negative correlation between pigeon numbers and woodland area (Wald = 6·78, d.f. = 1, P = 0·009). Woodland area did not significantly influence any of the other bird groups.

The overall composition of bird groups differed significantly between pheasant-managed and control woodlands (Λ = 0·926, F5,153 = 2·42, P = 0·04) (Fig. 1). A ranking matrix (Table 3) showed that relative to control woods, woodpigeons constituted a larger proportion of the bird community in pheasant woods. There was also a significant regional difference in the composition of bird groups (Λ = 0·879, F5,153 = 4·17, P = 0·001) with a higher proportion of woodpigeons in southern woods relative to eastern woods. There was no difference in the overall composition of bird groups between randomly and non-randomly selected woods (Λ = 0·979, F5,153 = 0·648, P = 0·66).

Figure 1.

Percentage compositions (mean ± 1 SE) of bird groups in pheasant managed woods (n = 81) and control woods (n = 78) during May–July 2004.

Table 3.  Ranking matrix for the relative proportions of bird groups in woodland interiors in pheasant-managed woods (n = 81) and control woods (n = 78) in southern and eastern England, summer 2004. Pheasant woods are represented by the rows and control woods by columns. + indicates the row group constituted a higher proportion than expected relative to the column; – indicates the opposite. A triple sign indicates that the difference was significant at P < 0·05
  1. Relative to control woods, pheasant woods ranked as follows, where the same letter indicates no significant difference: pigeonsa > othersa > warblersab > ground-feedersb > titsb > finchesb.

Pigeons+ + ++ + +++ + ++5
Others++ + ++4

Species richness did not differ between pheasant and control woods (Wald = 3·09, d.f. = 1, P = 0·08) but was higher in southern woods than eastern woods (mean species per wood ± 1 SE, 7·51 ± 0·25 and 6·95 ± 0·25, respectively, Wald = 7·27, d.f. = 1, P = 0·007). There was a significant effect of date category (Wald = 25·37, d.f. = 2, P < 0·001) and a negative relationship with woodland area within 1 km (Wald = 4·21, d.f. = 1, P = 0·04). The time of day that surveys were conducted also influenced species richness, with more species recorded in surveys started before 12·00 (Wald = 4·89, d.f. = 1, P = 0·03). Simpson's diversity index differed between regions (Wald = 7·05, d.f. = 1, P = 0·008), with a higher diversity index recorded in southern woods. Date was a significant factor, with a higher index recorded in earlier surveys (Wald = 17·00, d.f. = 2, P < 0·001). Wind was also a significant factor, with a lower index associated with higher wind categories (Wald = 6·48, d.f. = 2, P = 0·04).

For the bird groups where we found significant treatment effects, we conducted further REML analyses to determine the influence of vegetation characteristics on bird numbers. There was a negative correlation between total bird numbers and amount of tree canopy cover (Wald = 4·99, d.f. = 1, P = 0·03) (Fig. 2a). Region (Wald = 30·57, d.f. = 1, P < 0·001) was again a significant factor influencing total bird numbers. There was a significant negative correlation between amount of canopy cover and pigeon numbers (Wald = 5·24, d.f. = 1, P = 0·02) (Fig. 2b). Region was also a significant factor (Wald = 9·53, d.f. = 1, P = 0·002). There were positive correlations between numbers of warblers and the amount of vegetation (Wald = 4·93, d.f. = 1, P = 0·03) and the amount of R. fruticosus (Wald = 13·77, d.f. = 1, P < 0·001). There was also a negative correlation between warbler numbers and amount of tree canopy cover (Wald = 4·23, d.f. = 1, P = 0·04) (Fig. 2c).

Figure 2.

Relationships between birds numbers and tree canopy cover. (a) Total birds, (b) pigeons and (c) warblers.


REML analysis of deer browsing yielded a treatment × region effect (Wald = 4·42, d.f. = 1, P = 0·04). There was more evidence of deer browsing (% sampling stations with browsing recorded) in pheasant woods in the south and more in control woods in the east (southern region, pheasant woods: 22·5 ± 2·4%, control woods: 16·2 ± 1·8%, eastern region, pheasant woods: 18·7 ± 2·9%, control woods: 22·1 ± 2·4%). Date category (Wald = 52·63, d.f. = 2, P < 0·001) and woodland area (Wald = 6·27, d.f. = 1, P = 0·01) were also significant. Examination of the data for deer species seen indicated that muntjac deer Muntiacus reevesi Raf. was the most abundant deer species recorded in eastern woods and roe deer Capreolus capreolus L. the most abundant species in southern woods.


A general decline in the quality of the field layer of woodlands has been cited as an important factor contributing to the decline of a range of passerine species which depend on grassy and shrubby cover in woodland for nesting and feeding (Perrins & Overall 2001; Fuller et al. 2005). According to Fuller (2001), 61% of woodland passerines use the field or shrub layer for nesting, feeding or both activities. The two main factors responsible for the deterioration in the quantity and quality of the field layer are intensified deer browsing due to higher deer numbers and increased canopy closure due to lack of management (Fuller 2001; Gill & Beardall 2001). In our study, the differences in numbers of birds recorded between pheasant woods and control woods are likely to be due to differences in characteristics of the field layer. In pheasant-managed woods we recorded higher densities of vegetation and R. fruticosus at ground level.

Deer browsing can significantly alter the structure and composition of the field layer in a number of ways (Gill & Beardall 2001; Stewart 2001; Sage et al. 2004). In particular, it can lead to the removal of R. fruticosus resulting in an open understorey and an increase in grasses (Gill & Beardall 2001). In our study we found regional differences in the relative browsing pressure in pheasant-managed and control woods. There was more evidence of deer browsing in pheasant woods in the south and more in control woods in the east. This suggests that in the southern region pheasant management may be leading to increased deer numbers and hence additional browsing, whereas in the east where deer numbers are greater, the extent of browsing may be influenced by other factors such as size of woodland.

We found more R. fruticosus and more grasses in pheasant woods, suggesting that factors other than just deer browsing are important in determining field layer characteristics. We suggest that the differences in the field layer between pheasant-managed and control woods are due, in part, to differences in their relative shadiness. Morecroft et al. (2001) found that the amount of canopy cover and R. fruticosus cover were correlated inversely. In this study pheasant woods were more open, providing favourable light conditions for the growth of grasses, forbs and R. fruticosus. Opening up of the canopy (sky-lighting) to create flushing points for pheasants and to encourage growth of the understorey are two of the most common practices undertaken by game managers in woodland (Sage & Swan 2003). These are likely to have contributed to the higher levels of field layer cover recorded in this study. It is unclear from the results of this study whether active management in pheasant woods has led to the creation of a more open structure with higher levels of ground cover, or whether pheasant managers have simply chosen woods with a more open structure in which to release pheasants. However, as the selection criteria in this study required that pheasant woods had been managed for pheasants for at least 25 years, it is highly likely that managing the woodlands for pheasants has contributed to some degree to the vegetative and structural characteristics of the woods.

Shrub planting and coppicing in woodland is often cited as a common management technique to increase the attractiveness of woods for pheasants (Robertson 1992; Sage & Swan 2003). However, we found no differences in the shrub layer characteristics between pheasant-managed and control woods or in the diversity of shrubs within the woods. This may be because we restricted our surveys to the woodland interior and avoided the woodland edge, which is where shrub planting is often concentrated to maximize the benefits for pheasants (Woodburn 1991; Sage & Swan 2003). It is also possible that shrub management through planting or coppicing may have been undertaken in some pheasant woods but that regrowth has been prevented due to high deer browsing pressure (Fuller 2001; Fuller et al. 2005). This may be particularly relevant in the southern region, where there was significantly more deer browsing in the pheasant-managed woods.

Higher densities of warblers were observed in the pheasant-managed woods than the control woods. These species require dense low cover for nesting (Fuller 1995; Perrins & Overall 2001), which was more abundant in the pheasant woods. Indeed, we found a positive correlation between warbler numbers and vegetation cover. It is likely, therefore, that the higher levels of ground cover associated with pheasant-managed woods contributed to the higher number of warblers in those woods. The decline in active management in British woodlands is likely to be one of the main causes of the decline of woodland songbirds (Amar et al. 2006). Managing woodlands for pheasants often includes traditional woodland techniques including coppicing, ride management and skylighting to encourage growth of the field layer (Ludolf, Robertson & Woodburn 1989). It is likely that these aspects of woodland management for pheasants have a conservation benefit for migratory warblers.

There were no differences in numbers of passerine species which do not require the field layer for nesting (e.g. tits and finches) between pheasant and control woods. This is perhaps not surprising, considering that, with the exception of canopy cover, there were no differences in the structural composition of the woodlands above the field layer.

We recorded higher numbers of pigeons in pheasant-managed woods. They also constituted a higher proportion of the bird assemblage in pheasant woods relative to control woods. This difference is likely to be related to the presence of supplementary grain in the proximity of pheasant woods or game cover planted on land adjacent to woodland. It is also possible that structural characteristics of pheasant woods may provide more favourable roosting sites for woodpigeons. For example, a woodland with an open canopy may provide easier access into roosting sites than a woodland with a higher degree of canopy closure.

Previous research has highlighted potential negative effects of pheasants on ground flora inside release pens in ancient semi-natural woodlands (Sage, Ludolf & Robertson 2005). In our study vegetation surveys were designed to measure abundance of vegetation types and structural characteristics of the woodland rather than species composition of ground flora. Therefore, it was not possible to determine whether pheasant management led to changes in species composition of ground flora. Sage et al. (2005) found that species diversity inside release pens was lower and soil fertility higher than in areas surrounding the pen. More work is needed to document the species composition of the field layer in the wider woodland area outside the release pen. In addition, we restricted our study to the woodland interior and these results may not necessarily apply to the woodland edge zone which is a distinct habitat (Fuller 1995). The management of woodlands for game birds, especially pheasants, includes planting new woods as well as the retention and active management of existing woodlands (Tapper 2005). As such, game management has played an important role in shaping the lowland agricultural landscape in Britain (Cobham Resource Consultants 1997; Howard & Carroll 2001). This paper highlights some areas where there may be wider biodiversity gains through habitat management for game-bird hunting. However, this must be set in the context of other research which has highlighted potential negative impacts, e.g. Sage et al. (2005). Further research is required to understand fully the costs and benefits of hunting game birds on associated habitats and wildlife (Sutherland et al. 2006).


We are grateful to all the landowners and woodland managers who granted access to woods. Diane Ling and Mark Cunningham assisted with fieldwork and Nicholas Aebischer provided statistical advice. Peter Rothery, Simon Thirgood and two anonymous referees provided helpful comments on the draft manuscript. This work was funded by the Habitat Research Trust (HART) and The Game Conservancy Trust.