Present address: Centre for Environmental Sciences, School of Civil Engineering and the Environment, University of Southampton, Highfield, Southampton SO17 1BJ, UK.
Gap-crossing decisions of woodland songbirds in Scotland: an experimental approach
Article first published online: 8 AUG 2005
Journal of Applied Ecology
Volume 42, Issue 4, pages 678–687, August 2005
How to Cite
CREEGAN, H. P. and OSBORNE, P. E. (2005), Gap-crossing decisions of woodland songbirds in Scotland: an experimental approach. Journal of Applied Ecology, 42: 678–687. doi: 10.1111/j.1365-2664.2005.01057.x
- Issue published online: 8 AUG 2005
- Article first published online: 8 AUG 2005
- Received 6 October 2004; final copy received 23 March 2005 Editor: Rob Freckleton
- forest habitat networks;
- habitat fragmentation;
- mobbing behaviour
- 1Recent declines in woodland birds in Britain have been linked to increasing habitat fragmentation. To understand the effects of fragmentation, data on avian dispersal across woodland gaps are essential but often lacking.
- 2We used song thrush Turdus philomelos mobbing calls to attract songbirds across gaps ranging from 5 to 120 m in width and along comparable woodland edges. Responses were modelled against distance using generalized linear models. Such models have clear applied value for connecting fragmented landscapes.
- 3We also calculated response indices and compared these with bird morphology. The gap-crossing results were applied to a real landscape in central Scotland and landscape metrics were calculated to judge how perception of habitat connectivity varies interspecifically.
- 4The chaffinch Fringilla coelebs and the robin Erithacus rubecula both responded more readily across gaps than through woodland. There was no difference between gap and edge response for the coal tit Parus ater, while the goldcrest Regulus regulus responded more readily along edges than across gaps. Maximum gap-crossing distances ranged from 46 m (goldcrest) to 150 m (chaffinch).
- 5There was a positive linear trend between mass of bird and the difference in the maximum response for gap and control experiments. Likewise there was a positive curvilinear relationship between wing area and the difference in probability of response between gap and control experiments at 50 m. These results may be interpreted in terms of manoeuvrability and ability to escape avian predation.
- 6For the central Scotland landscape, the perceived number of patches in the landscape decreased exponentially with increasing gap-crossing distance, while the median patch size and mean patch fractal dimension increased linearly with gap-crossing distance.
- 7Synthesis and applications. Our results show that an experimental approach using playback can be used to obtain data on avian gap crossing and the results applied to real landscapes to visualize interspecific differences in habitat perception. This has practical management applications, especially for designing forest habitat networks to maximize avian biodiversity, and potentially could help reverse the recent declines in woodland birds.
Woodland birds have declined in Britain by 22% since the mid-1970s (Gregory et al. 2003), possibly in part as a result of increasing habitat fragmentation (Bellamy et al. 2000). Fragmentation leads to a loss of habitat area and an increasing distance between remaining habitat patches (Andrén 1994; Opdam et al. 1994). The point at which the effects of fragmentation become greater than that purely the result of loss of habitat area will be species specific, depending on dispersal ability (Andrén 1994; With 1997). As birds are highly vagile, it has been suggested that they may be insensitive to habitat fragmentation (Margules, Higgs & Rafe 1982; Ambuel & Temple 1983). Physically they are able to cross gaps of many kilometres in width, and energetic cost differentials will be relatively small for gaps of up to 1 km. However, a psychological reluctance may restrict the gap widths that they are willing to cross (Kirby 1995; Grubb & Doherty 1999), possibly because of the birds’ perception of predation risk (Lima & Dill 1990).
There is increasing evidence that habitat extent and connectivity are important for at least some bird species, and numerous observational studies have suggested that wood size and isolation are important determinants of bird species richness (Howe 1984; Opdam, Van Dorp & Ter Braak 1984; Opdam, Rijsdijk & Hustings 1985; McCollin 1993; Léscourret & Genard 1994). Likewise, studies based on ringing recoveries and the tracking of marked birds have found that fragmentation can inhibit dispersal (Matthysen, Adriaensen & Dhondt 1995; Haas 1995; Grubb & Doherty 1999). Two categories of gap crossing can be distinguished from these studies. Dispersal gap crossing occurs between natal areas and breeding sites, is relatively rare, and often covers large areas. Home range gap crossing, on the other hand, is relatively small scale and connects habitat patches within the home range of a bird.
Data on willingness to cross gaps are unavailable for most bird species, especially as tracking animals individually is both technologically and labour intensive (Rushton, Ormerod & Kerby 2004), yet a proper understanding of the scale at which the landscape is perceived is essential for identifying when species will become sensitive to fragmentation. Such information is especially pertinent in Britain as woodland planting grants are increasingly targeted to areas where they will contribute to the creation of forest habitat networks. New woodland contributes to a forest habitat network where it enlarges and reconnects existing native woodland remnants at either a macro- or a micro-scale (Forestry Commission 2001). Current guidelines recommend at least 30% woodland cover in the landscape, beyond which diminishing returns in ecological benefits occur. Gaps between woodland fragments of up to 30 m are acceptable, but an acknowledgement is made that the gap widths that species will cross is crucial (Forestry Commission 2001). Only limited experimental data underlie these guidelines.
A new technique for studying bird dispersal was suggested by Desrochers & Hannon (1997) using the playback of avian mobbing calls to elicit directional movements in birds across different habitat configurations in the landscape. Birds carry out mobbing in response to a stationary predator, and several characteristics of mobbing behaviour make it an effective tool for studying bird movement through the landscape. First, mobbing calls cover a wide audio-frequency range, making them highly locatable (Shalter & Schleidt 1977; Klump & Shalter 1984). Secondly, mobbing calls attract both conspecifics and heterospecifics (Curio 1978; Vieth, Curio & Ernst 1980; Hurd 1996; Desrochers & Hannon 1997), possibly because of similarities between bird species (Marler 1955). Thirdly, mobbing behaviour occurs all year round, at least in some species (Hinde 1952; Curio 1978).
In this study we used an adaptation of the Desrochers & Hannon (1997) method to study gap crossing by woodland songbirds in Scotland, and then visualized how species perceive the landscape by applying the results in a GIS. We addressed the following questions. (i) What is the effect of narrow gaps (up to 120 m) in woodland cover on the movement of songbirds? (ii) How do responses to gaps vary between species, and are these differences related to bird morphology? (iii) What is the potential of the playback methodology for obtaining information on avian dispersal behaviour? (iv) What are the implications for the creation of forest networks to minimize the impacts of fragmentation?
The habitat was coniferous plantation with a mosaic of spruce Picea spp., larch Larix spp., pine Pinus spp. and Douglas fir Pseudodtsuga menziesii Mirbel, Franco at East Loch Lomond, Loch Ard Forest and Achray Forest in the west of Scotland. Patches of mature, oak-dominated Quercus spp. semi-natural woodland occurred throughout the forests. The coniferous woodland varied in age from newly replanted areas to stands of around 70 years old. The dominant tree age was around 50 years. Thinning had not been carried out so stands were dense and generally had little or no ground vegetation. Gaps in the woodland cover were created by forest roads, powerline corridors and clearfelling. There was therefore some variation in habitat within gaps, but habitat was not consistently different between gaps of different widths and was therefore unlikely to bias the results. Habitat variation was unavoidable because of the requirement for sufficient independent locations for experiments. Most of these gaps were less than 200 m wide, although some clearfell areas exceeded this, giving gaps as large as 1 km in width.
Experiments were carried out between 9 February and 17 March 1999, 11 July and 27 August 1999 and 6 May and 11 August 2000. A recording of song thrush Turdus philomelos Brehm mobbing calls was used to attract birds across gaps in the woodland cover ranging from 5 m to 120 m in width. The song thrush call was chosen because the recording was of good quality and elicited a consistent response from a range of species. Mobbing calls were audible to us up to 190 m across a gap and up to 160 m through woodland, an audible range exceeding the widest woodland gap studied. Sites of playback experiments were at least 400 m apart in the landscape. Based on the audible range of the equipment this was sufficient to ensure that experiments were independent. Each location was only used once within any field season. Experiments were carried out under conditions of no rain or strong wind, between half an hour after sunrise and 16:00 h. By performing the experiments in a relatively uniform habitat, avoiding features such as streams (which produce additional background noise) and only collecting data under a limited range of weather conditions, sound attenuation should have been relatively constant.
Upon arrival at an experimental site the playback equipment was set up at a woodland edge, but not switched on. A 2-min settling period was allowed and was followed by a 10-min period of observation during which a bird point count was carried out and any gap crossing or bird movements away from the woodland edge were recorded. Playback was then commenced and lasted 10 min. Two types of experiments were performed: gap experiments and control experiments. Control experiments were important because the probability of response declines naturally with distance because of the effects of sound attenuation and degradation on the signal. Also birds may not regard a predator (which mobbing calls indicate) as a threat over large distances so there would be no need to mob. The control experiments provided a measure of this natural rate of decline. In both gap and control experiments any bird moving to within 5 m of the speaker was deemed to be responding. Response was unambiguous because it was almost always accompanied by aggressive posturing and calls from the birds.
Desrochers & Hannon (1997) identified individual birds within 10 m of the woodland edge prior to starting the playback and then followed these to determine whether they responded or not. This method proved unworkable in our study area as individual birds could not be kept constantly in view within the dense conifer canopy. We therefore adapted the methodology. In gap experiments, the number and species of any birds responding at the speaker and crossing the gap was recorded. Experiments where the minimum response at the speaker was fewer than five birds were rejected to avoid bias. Usually there was an identifiable reason for this lack of response, including a sudden deterioration in the weather conditions, loss of power in the playback equipment and the presence of an aerial predator. We carried out at least 10 valid gap-crossing experiments for each 10-m increment of gap width by choosing from available forest gaps, except for the width 90–100 m where insufficient sites were available. A total of 182 gap-crossing experiments was carried out, of which 138 were classed as valid.
Control experiments were designed to measure the response of the birds through woodland. The playback equipment was set up in the woodland edge as in gap-crossing experiments, with the observer positioned in the open habitat away from the edge. Once playback commenced, the observer walked at a steady pace along the woodland edge until a stationary bird was located within the woodland. This bird was then observed for 1 min. If it showed signs of directional movement towards the speakers, a marker was placed opposite its original location and it was followed to see if it genuinely was approaching the speaker. Distances between the original location of control birds and the speaker were measured using the markers once the playback had finished. Birds observed once they were already moving towards the speaker were not counted. Control experiments were designed so that approximately equal lengths of observational time during the playback would be carried out at all distances between 0 and 120 m along the wood edge. Thus there would be an equal probability of detecting a responding bird at all distances within this distance range.
There were some limitations to the control method. A stationary bird could have responded over some distance between the time at which playback commenced and the time at which it was observed. Thus the estimate of maximum response distances may be quite conservative. Attempts were made to identify birds in the woodland edge prior to playback, but this proved unworkable. Playback stimulated the birds to posture and call even prior to them approaching the playback equipment and this meant that they became easier to observe and follow. A total of 123 control experiments was carried out. Data were obtained from 85 of these. There was no confounding effect of observer because all data were collected by the same individual.
Sufficient data for analysis were obtained for four species: chaffinch Fringilla coelebs L., goldcrest Regulus regulus L., coal tit Parus ater L. and robin Erithacus rubecula L. Data were reduced to presence or absence of gap crossing for both the periods before and after playback commenced. This was to take account of the fact that the total number of individuals of each species gap crossing could be sensitive to the population density in the area, which was unknown. The proportion of experiments in which gap crossing occurred for each species was calculated separately for non-playback and playback time periods for 10-m gap increments. It was assumed that there were no significant effects of weather, time of day or time of year. Formal testing of this assumption was not possible but because gap and control experiments were randomly distributed with respect to time of day and year, no systematic biases should have been introduced into the data.
The relationship between probability of response and gap distance was modelled for the playback experiments in S-Plus (Mathsoft 1999), using generalized linear modelling with a logistic link function assuming a binomial error distribution. The explanatory power of the models was assessed using the adjusted D2, where the adjustment was calculated using an adaptation of Weisberg's (1980) adjusted R2 measure. Data for probability of response before the playback were too sparse to allow formal modelling, so these were plotted and best-fit lines were added for illustrative purposes only.
The total number of each species responding along woodland edges was collated for 10-m distance increments away from the speaker. Where more than one individual had responded simultaneously along the woodland edge, starting from the same location, this was counted as a single registration to avoid pseudoreplication, as the response of one individual may influence others (Kroodsma 1986). Poisson regression models were fitted using distance as the predictor. The count data and fitted values were then rescaled to allow comparison of the response through woodland with that across gaps. This involved extrapolating the regression line for the gap-crossing data for each species to a distance of 0 m and rescaling the control data relative to this probability of response value.
To facilitate comparison of gap and control responses between the four species, the difference in probability of response for a distance of 50 m from the speaker was calculated. In addition, the distance at which the probability of response was effectively zero (taken as the point at which probability reached 0·05) was determined for both gap and control experiments and the difference between these distances was calculated for each species. These measures were plotted against bird mass, absolute and relative measures of wing span and area, and wing loading. Although these graphs had only four data points on them, they were useful to suggest hypotheses worthy of further investigation.
To demonstrate how gap-crossing data may be used to interpret how species perceive the connectivity of a landscape, a 10 × 10-km area was selected between Edinburgh and Glasgow in central Scotland. GIS data of woodland distribution digitized from aerial photographs were available for this area. The selected landscape contained a total wooded area of 838·45 ha (8·4% cover), of which 567·81 ha (5·7%) was broad-leaved woodland and 270·65 ha (2·7%) was coniferous or mixed broad-leaved and coniferous woodland. The total number of patches in the landscape, the median patch size and the mean patch fractal dimensions were calculated for this area assuming non-contiguous fragments to be entirely separate entities. Mean patch fractal dimension is a measure of shape complexity, taking a value between one and two where a higher value is indicative of greater shape complexity (McGarigal & Marks 1994). These values were then recalculated for each bird species, where individual fragments were classed as connected when they were less than the maximum gap-crossing distance apart. The maximum gap-crossing distance was determined from the generalized linear models for gap crossing under playback conditions for each species. All calculations were carried out in ArcView 3.2 using Patch Analyst (Elkie, Rempel & Carr 1999).
birds in gaps under non-playback conditions
The chaffinch was the most frequently observed of the four species crossing gaps when playback was not being used. Chaffinches crossed gaps of up to 120 m in width. The proportion of experiments with observed crossings for robins and coal tits was less than 0·25 for gaps of up to 10 m in width. Maximum observed gap-crossing distances were less than 50 m for both these species (Fig. 1). No gap-crossing activity was observed for goldcrests.
A comparison of Fig. 1 with Fig. 2 shows that the probability of gap crossing over any distance was much lower under non-playback conditions compared with playback conditions for all species. Likewise the estimated maximum distance of gap crossing was greater for playback experiments than for non-playback time periods.
Occurrences of birds within gaps when no directional movement was occurring were also recorded (Table 1). As with the gap-crossing result under non-playback conditions, the goldcrest was never observed in the gap away from the woodland edge and the coal tit only very rarely so (in less than 1% of counts, which represented only two sightings). Chaffinches and robins used gap habitats with a low frequency.
A total of 1555 birds of 17 different species was recorded at the speaker in the gap-crossing and control experiments. This figure is unlikely to be a completely accurate representation of total response because, in spite of the distance between experimental sites, some birds may have responded in more than one experiment, while in control experiments the speaker was not constantly in view of the observer and some responding individuals may not have been counted. Approximately 90% of recorded respondents comprised the chaffinch, coal tit, goldcrest and robin, and these all responded in more than 45% of valid experiments. Almost 50% of birds responding were chaffinches, which is unlikely to be directly proportional to their abundance in the environment.
The decline in probability of response with distance was significant for both gap and control experiments for the four species analysed. Distance explained more than 50% of variation in response for all models except the goldcrest control model (Table 2). Goldcrest control data were relatively sparse, because this was the hardest of the four species to follow as it moved towards the speaker. This could explain the relatively low adjusted D2 in this instance.
|Species||Gap crossing full binomial model||Control Poisson model|
|B||F||P||Adj. D2||B||F||P||Adj. D2|
The chaffinch and the robin, the two species with the largest maximum gap-crossing distances under non-playback conditions (Fig. 1), were also more likely to respond across gaps than through woodlands for all distances studied (Fig. 2a,b). There was no difference in probability of response for gaps and woodland for the coal tit (Fig. 2c). In contrast, the goldcrest responded more readily through woodland than across gaps for all distances (Fig. 2d). The estimated maximum gap-crossing distance was greatest for the chaffinch (150 m) and least for the goldcrest (46 m). The results for the robin and the coal tit fell between these two extremes (Table 3).
|Predicted maximum gap-crossing distance (m) under playback||150||60||92||46|
|Predicted maximum woodland response distance (m) under playback||83||24||113||83|
|Difference between predicted maximum gap and woodland response distance (m) under playback||67||36||−21||−37|
|Predicted maximum gap-crossing distance (m) with no playback||120||42||29||0|
|Difference between predicted probability of response at 50 m for gap and control playback experiments||0·399||0·062||−0·003||−0·061|
gap crossing behaviour in relation to bird morphology
Although caution must be exercised when data are available for only four species, possible trends were observed in plots of response indices against two morphological measures. There was a positive linear trend between mass of bird and the difference in the maximum response for gap and control experiments (Fig. 3). Likewise there was a positive curvilinear relationship between wing area and the difference in probability of response between gap and control experiments at 50 m (Fig. 4).
landscape spatial statistics
Figure 5 illustrates, for a single initial fragment, how increasing numbers of woodland fragments will be perceived as part of a single patch based on an increasing willingness to cross wider open areas. The total number of patches, mean patch size and mean patch fractal dimension of the central Scotland landscape for the maximum gap-crossing distance for each species under playback conditions are given in Table 4. Plotting these relationships suggested that the number of patches in the landscape decreases exponentially with increasing gap-crossing distance (Fig. 6), while the median patch size and mean patch fractal dimension increase linearly with gap-crossing distance (Figs 7 and 8, respectively). The scarcity of the data meant that these relationships were not formally modelled.
|Species||Distance (m)||NumP||MedPS (ha)||MPFD|
interspecific differences in gap-crossing behaviour
Clear interspecific differences existed in the willingness of woodland birds to cross gaps, supporting the argument for a species-centred approach in looking at landscape pattern (Andrén 1994; Kirby 1995; With 1997; With, Gardner & Turner 1997). Of the four species studied the goldcrest was the most inhibited by breaks in the woodland cover. It was never recorded in gaps under non-playback conditions, responded to playback much more readily through woodland than across gaps and had the shortest maximum gap-crossing distance. In contrast, there was no difference in the probability of response across gaps relative to the probability of response through woodland for the coal tit, while the chaffinch and robin responded more readily across gaps than through woodland. The control experiments were designed to measure the decline in response with distance expected because of sound attenuation. Based on this it could be argued that for the coal tit no evidence was found of a behavioural reluctance to cross gaps of increasing width. A similar interpretation could be made for the chaffinch and the robin taking account of the fact that no correction was made for the excess sound attenuation as a result of the vegetation through woodland. Consequently the audible range of the playback was greater across gaps than through woodland and this could account for the difference in the response curves for these two species. However, this interpretation of the data is unlikely to represent the whole picture. All response curves differed interspecifically and, with the exception of the coal tit, between woodland and gaps. Audibility curves, however, tend to be very similar between bird species, with oscines showing even less variability than nonoscines (Dooling 1982). Furthermore, none of the response curves precisely followed the theoretical sound attenuation curve for the study area, even allowing for a shift in the curve for different audible thresholds (Fig. 9). In all cases the maximum gap distance was within the audible range of the observer, and human audibility thresholds are lower for all sound frequencies than those of oscine birds (Dooling 1982). Therefore, while a sound attenuation effect may account for some of the pattern in the data, other factors are also likely to be important.
An alternative explanation for the interspecific differences in gap crossing is suggested by the relationships between response indices and morphological parameters. As bird mass increased, species responded less readily through woodland and more readily across gaps. Likewise as wing area increased the probability of a response across gaps relative to through woodland increased in a curvilinear way. It is possible that the larger birds with bigger wings were less manoeuvrable within woodland and better adapted to flight in a more open habitat. The coal tit and goldcrest are adapted to low speed, manoeuvrable flight that is advantageous within woodland (Norberg 1979; Rayner 1979). The robin and chaffinch may be able to fly faster and be more manoeuvrable in the open, and these parameters determine the probability of a bird evading an aerial raptor attack (Rudebeck 1950; Newton 1986; Cresswell 1993). Consequently the perceived risk of crossing wider habitat gaps may be less for robins and chaffinches.
Additional studies are also needed to investigate the hypothesis that the relative response differences between woodland and gaps relate to morphological adaptations for flight and how this affects the ability to evade a predator. Four species represent too small a sample size to have confidence in the form of the relationships. The strongest support could be derived from investigating species that may be expected to deviate from the relationship in predictable ways based on other morphological parameters that could affect their manoeuvrability and speed within or outside woodland.
interpretation of the results for forest habitat networks
It is not possible to say whether the maximum gap-crossing distances recorded under playback conditions represent the true maximum gap distances that individuals are prepared to cross within their home range. Mobbing is a behavioural response to the supposed presence of a predator and entails costs in terms of time and energy expenditure (Curio 1978) and the risk of being killed or injured (Curio & Regelmann 1985). Under different motivation it is possible that greater distances would be crossed. An unpublished study (F. McColm, personal communication), where food was used as an attractant, indicated that willingness to move away from woodland edges does vary under different stimuli. Likewise the ranging behaviour of individuals can vary at different times of the breeding season as motivation changes (Peach et al. 2004). However, the preference of the goldcrest for moving through woodland rather than across gaps suggests that continuous networks are important for at least some bird species. The maximum gap-crossing distance of this species (the one most reluctant to cross gaps) exceeded the Forestry Commission (2001) guideline of 30 m as an acceptable distance between patches in a forest habitat network. Thus, for the four study species, a forest habitat network created based on the guidelines would be perceived as connected. The current scientific basis for this recommendation is sparse and this finding provides some justification, especially as the goldcrest is the smallest European bird. If the hypothesis relating morphology and gap crossing has credence, this guideline may be more widely applicable.
The exercise in calculating landscape metrics demonstrated how gap-crossing information could be used to identify how different species perceive landscape connectivity. There was an exponential decline in the number of blocks of unconnected patches in the landscape with increasing gap-crossing ability. This could indicate that forest patches in the central Scotland forest were relatively randomly distributed, as the relationship mirrors that found by Andrén (1994). His study demonstrated an exponential increase in the distance between patches as habitat area was reduced in a randomly patchy landscape with less than 20% habitat cover. In the Central Scotland study area the cover was only about 8%. If the forms of Andrén's (1994) relationships are relatively constant it would be possible to estimate habitat connectivity for any landscape for any of these species based on the percentage cover of habitat in the landscape and how patches were distributed (e.g. random, aggregated or over-dispersed). Guidelines for optimal locations for additional habitat creation to bring the landscape total up to the 30% cover recommended by the Forestry Commission (2001) could also be developed.
evaluation of the methodology
This study has demonstrated that a modified form of Desrochers & Hannon's (1997) mobbing call playback methodology may be used to investigate willingness of birds to cross gaps in dense coniferous woodland. This is the first attempt to our knowledge to replicate their methodology to study gap-crossing behaviour. The main limitation to the modified method is that it is highly labour intensive, requiring many field work hours to collect a relatively small amount of data. Desrochers & Hannon's (1997) method was also labour intensive, but less so. In their gap-crossing experiments each playback could potentially contribute more than one observation to the data set, whereas in the modified method at least 10 playback sessions were required to generate a single data point. This required an extensive study area so that sufficient independent sites for playback were available and meant that data had to be collected over a wider time period than was desirable. Motivation to mob may vary at different times of year (Smith & Graves 1978), but this should not have biased the data as experiments that did not achieve a minimum level of response were discounted. Some reduction in the field work hours could be made by not measuring gap-crossing responses for every 10-m gap increment. If sufficient gap distances were measured it would still be possible to determine the shape of the relationship between probability of gap crossing and gap distance.
Using a modification of Desrochers & Hannon's (1997) method, we found clear interspecific differences in willingness to cross gaps relative to willingness to move through woodland. These results cannot be fully explained by sound attenuation and we suggest that a better explanation lies in morphological adaptations to flight and how these affect ability to evade a predator. The gap-crossing results can be applied to real landscapes to gain an idea of how the perception of habitat connectivity may vary with different species. This has practical management application, especially for designing forest habitat networks that attempt to reconnect fragmented forest landscapes.
We should like to thank the following organizations for their assistance in this work: the Forestry Commission for allowing access to their woodlands for fieldwork and for providing forest maps; Highland Birchwoods for providing the GIS data for central Scotland; and the National Sound Archive for providing the mobbing call recording. We also thank André Desrochers for helpful comments on his playback methodology. H. P. Creegan was funded by a University of Stirling studentship.
- 1983) Area-dependent changes in the bird communities and vegetation of Southern Wisconsin forests. Ecology, 64, 1057–1068. & (
- 1994) Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos, 71, 355–366. (
- 2000) Variation in the relationship between numbers of breeding pairs and woodland area for passerines in fragmented habitat. Ecography, 23, 130–138. , , & (
- 1993) Escape responses by redshanks Tringa totanus, on attack by avian predators. Animal Behaviour, 46, 609–611. (
- 1978) The adaptive significance of avian mobbing. I. Teleonomic hypotheses and predictions. Zeitschrift fur Tierpsychologie, 48, 175–183. (
- 1985) The behavioural dynamics of great tits (Parus major) approaching a predator. Zeitschrift fur Tierpsychologie, 69, 3–18. & (
- 1997) Gap crossing decisions by forest songbirds during the post-fledging period. Conservation Biology, 11, 1204–1210. & (
- 1982) Auditory perception in birds. Acoustic Communication in Birds. Volume 1. Production, Perception and Design Features of Sound (eds D.E.Kroodsma & E.H.Miller), pp. 95–130. Academic Press, London, UK. (
- 1999) Patch Analyst User's Manual. Ontario Ministry of Natural Resources, Northwest Science and Technology, Thunder Bay, Ontario, Canada. , & (
- Forestry Commission (2001) Guidance Note 20 December 2001: Forest Habitat Networks. Forestry Commission, Edinburgh, UK. http://www.forestry.gov.uk/pdf/gnote.pdf/$file/gnote.pdf[accessed on 2 February 2004].
- 2003) The State of the UK's Birds 2002. RSPB, BTO, WWT & JNCC, Sandy, UK. , , , , , , & (
- 1999) On home-range gap crossing. Auk, 116, 618–628. & (
- 1995) Dispersal and use of corridors by birds in wooded patches on an agricultural landscape. Conservation Biology, 9, 845–854. (
- 1952) The behaviour of the great tit (Parus major) and some related species. Behaviour, 2 (Supplement), 1–201. (
- 1984) Local dynamics of bird assemblages in small forest habitat islands in Australia and North America. Ecology, 65, 1585–1601. (
- 1996) Interspecific attraction to the mobbing calls of black-capped chickadees (Parus atricapillus). Behavioral Ecology and Sociobiology, 38, 287–292. (
- 1995) Rebuilding the English Countryside: Habitat Fragmentation and Wildlife Corridors as Issues in Practical Conservation. English Nature, Peterborough, UK. (
- 1984) Acoustic behaviour of birds and mammals in the predator context. Zeitschrift fur Tierpsychologie, 66, 189–226. & (
- 1986) Design of song playback experiments. Auk, 103, 640–642. (
- 1994) Habitat, landscape and bird composition in mountain forest fragments. Journal of Environmental Management, 40, 317–328. & (
- 1990) Behavioural decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology, 68, 619–640. & (
- 1993) Avian distribution patterns in a fragmented wooded landscape (North Humberside, UK): the role of between-patch and within-patch structure. Global Ecology and Biogeography Letters, 3, 48–62. (
- 1994) Fragstats Spatial Pattern Analysis Program for Quantifying Landscape Structure: Version 2.0. Forest Science Department, Oregon State University, Corvallis, OR. & (
- 1982) Modern biogeographic theory: are there any lessons for nature reserve design. Biological Conservation, 24, 115–128. , & (
- 1955) Characteristics of some animal calls. Nature, 176, 6–8. (
- Mathsoft (1999) S-PLUS 2000 User's Guide. Data Analysis Products Division, MathSoft, Seattle, WA.
- 1995) Dispersal distances of nuthatches, Sitta europaea, in a highly fragmented forest habitat. Oikos, 72, 375–381. , & (
- 1986) The Sparrowhawk. Poyser, Calton, UK. (
- 1979) Morphology of the wings, legs and tail of three coniferous forest tits, the goldcrest and the treecreeper in relation to locomotor pattern and feeding station selection. Philosophical Transactions of the Royal Society of London B, 287, 131–165. (
- 1994) The landscape ecological approach in bird conservation: integrating the metapopulation concept into spatial planning. Ibis, 137 (supplement 1), 139–146. , , & (
- 1985) Bird communities in small woods in an agricultural landscape: effects of area and isolation. Biological Conservation, 34, 333–352. , & (
- 1984) The effect of isolation on the number of woodland birds in small woods in the Netherlands. Journal of Biogeography, 11, 473–478. , & (
- 2004) Habitat selection by song thrushes in stable and declining farmland populations. Journal of Applied Ecology, 41, 275–293. , , , , , , & (
- 1979) A new approach to animal flight mechanics. Journal of Experimental Biology, 80, 17–54. (
- 1950) The choice of prey and modes of hunting of predatory birds with special reference to their selective effect. Oikos, 2, 65–88. (
- 2004) New paradigms for modelling species distributions? Journal of Applied Ecology, 41, 193–200. , & (
- 1977) The ability of barn owls, Tyto alba, to discriminate and localize avian alarm calls. Ibis, 119, 22–27. & (
- 1978) Some factors influencing mobbing behavior in barn swallows (Hirundo rustica). Behavioural Biology, 23, 355–372. & (
- 1986) Flight cost of a small passerine measured using doubly labelled water: implications for energetic studies. Auk, 103, 169–180. & (
- 1980) The adaptive significance of avian mobbing. III. Cultural transmission of enemy recognition in blackbirds: cross-species tutoring and properties of learning. Animal Behaviour, 28, 1217–1229. , & (
- 1980) Applied Linear Regression. Wiley, New York. (
- 1997) The application of neutral landscape models in conservation biology. Conservation Biology, 11, 1069–1080. (
- 1997) Landscape connectivity and population distributions in heterogeneous environments. Oikos, 78, 151–169. , & (