• field selection;
  • pesticide residues;
  • repellency;
  • risk assessment;
  • wildlife poisoning


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. Pesticides applied to seeds before sowing may present a high risk to seed-eating birds. Some of these chemicals are highly toxic but are avoided by captive birds in laboratory tests. However, evidence for birds avoiding them in the field is lacking. The objective of this study is to investigate avoidance and other factors affecting risk of poisoning, using the model of wood-pigeons Columba palumbus L. feeding on winter-drilled cereal seed treated with the organophosphorus insecticide fonofos.

2. Wood-pigeons feeding on arable fields in Cambridgeshire, UK, were counted once or twice per week over four winters from 1992 to 1996. In autumn, newly drilled cereal fields are shown to be preferred to alternative crops such as stubble and ploughed fields and root crops. In spring, newly drilled cereal fields are sometimes preferred to older cereal fields, stubble, ploughed fields and drilled onions.

3. Fewer fields drilled with fonofos-treated cereal seed were used by wood-pigeons than untreated (without fonofos) fields for the first week after drilling only. The extent of the avoidance reaction was related to the concentration of fonofos on the seed left exposed on field surfaces.

4. The concentration of fonofos on newly drilled seed exposed on field surfaces (mean of 241 mg kg–1) was much lower than expected from the approved application rate (1080 mg kg–1). Residues decayed linearly at a rate of 2% per day after drilling and varied widely between fields.

5. The density of seed left exposed after drilling also varied widely between fields, being greater on headlands (field edges) than the main field and declining at a rate of 2% per day after drilling.

6. Of 61 wood-pigeons shot within the study area over three winters, 26 had been feeding on cereal seed and of these, six contained fonofos residues at levels similar to poisoned birds submitted to the MAFF Wildlife Incident Investigation Scheme.

7. In conclusion, fonofos-treated cereal seed may be partially avoided in the wild. Avoidance fails to prevent poisoning under some (rare) conditions, possibly when seed density, concentrations of fonofos and levels of hunger are all above average. The implications for risk assessment are discussed.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Populations of several European farmland birds have declined over the last 20 years (Petersen 1994; Saris et al. 1994; Fuller et al. 1995). These declines have been linked to changes in farming practice; in particular increased pesticide usage, the switch to autumn-sown crops and the consequent lack of winter stubble fields with surface grain, the loss of hedgerows, changes in the relative areas of tillage or grassland, and more complete harvesting and improved storage techniques (Donald & Forrest 1995; Feare 1989; Inglis et al. 1990; Tucker 1992; Wilson, Taylor & Muirhead 1996). Some recent studies have focused on whether the effects of pesticides have contributed to these declines (Evans, Wilson & Browne 1994; Campbell et al. 1997) but as yet the results are inconclusive (except for grey partridge Perdix perdix;Campbell et al. 1997).

As a result of public concern for the environment and health, some European countries have adopted a policy of minimizing pesticide inputs (for example Sweden; Ekstrom & Bernson 1995). The environmental and social costs of pesticides have to be balanced against their benefits in terms of reduced crop losses to pests, diseases and weeds (Bowles & Webster 1995). There is therefore a need for policy makers to identify effective pest management practices with the least detrimental impact (Levitan, Merwin & Kovach 1995) so that farmers can use low-cost relatively environmentally friendly chemicals within integrated crop management systems (OECD 1996b).

Pesticide risk assessments are currently undertaken in many countries in order to estimate the risk of new and existing pesticides to wildlife. The assessment generally involves comparing toxicity and exposure. Toxicity is extrapolated from laboratory surrogates. For birds, the first estimate of exposure often assumes that individuals obtain all their daily food requirements from pesticide-contaminated food. Many pesticides can be shown to be avoided by birds under laboratory conditions, which should reduce exposure and hence risk, but there is little evidence for avoidance in the field (OECD 1996a). The term avoidance is used here in a general sense, to mean behaviour resulting in a reduction in the amount ingested, for whatever reason. There is therefore a need to determine whether avoidance does reduce the exposure of birds to pesticides in the field: (i) to help regulators interpret existing palatability data; and (ii) to assist the development of improved test methods for risk assessment.

This study was part of a larger project designed (i) to compare predicted and actual risks to birds from seed treatments; (ii) to identify factors affecting risk; and (iii) if necessary, to develop improved methods for risk assessment. The application of pesticides as seed treatments is thought to reduce the exposure of non-target organisms when compared to equivalent sprays or granular formulations (Dewar & Asher 1994). The risk to granivorous birds, however, may be substantially increased. In fact, several cases of bird mortality caused by organophosphorus cereal seed treatments have been reported (Stanley & Bunyan 1979; Greig-Smith 1987; Fletcher, Hunter & Barnett 1995) and can be very significant. Stanley & Bunyan (1979) concluded that the deaths of pink-footed geese Anser brachyrhynchus and greylag geese A. anser due to carbophenothion represented ‘a significant proportion of the world population’.

Wood-pigeons Columba palumbus L. were studied in an area of Cambridgeshire fenland, UK, where the organophosphorus seed treatment fonofos (O-ethyl S-phenyl (RS)-ethylphosphonodithioate) was used to combat the wheat bulb fly Delia coarctata (Fallén). The manufacturers have since withdrawn this chemical from the market. This model system was chosen because:

1. wood-pigeons are very common in England;

2. wood-pigeons are among the species most frequently involved in poisoning incidents caused by seed treatments in the UK (Wildlife Incident Investigation Scheme data: Fletcher et al. 1991; Fletcher et al. 1992; Fletcher, Hunter & Barnett 1992; Fletcher, Hunter & Barnett 1995; Fletcher et al. 1996);

3. fonofos was currently the most widely used organophosphorus seed treatment (although its use was concentrated in a few geographical areas; M. Thomas, MAFF Pesticide Usage Survey, unpublished data); and

4. wood-pigeons are known to feed on newly sown cereal seed and the hazard posed by seed-dressings has been examined in previous studies (Murton & Vizoso 1963; Cooke 1988). Captive birds, however, can be shown to avoid fonofos-treated cereal seed (Pascual & Hart 1997; Fryday, Chandler-Morris & Hart 1998).

This paper reports the results of a field study on the use of agricultural fields by wood-pigeons in the four winters between 1992 and 1996, to investigate avoidance and other factors affecting the risk of poisoning.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The study took place in a 30-km2 area of Cambridgeshire fenland near Ramsey (52°27′N, 0°13′W) in England. The study site was chosen because of the relatively widespread use of fonofos to control wheat bulb fly. The fields are typically high organic matter peat soils in an artificially drained low-lying area and are intensively cropped, mainly with sugar beet Beta vulgaris, potatoes Solanum tuberosum and wheat Triticum sp. (Caldwell & Richardson 1975; Mason 1975). The ecology of wood-pigeons in the study area is described in Murton, Isaacson & Westwood (1963).


In September 1992, a route through the study area was chosen from which 66 arable fields could be easily observed from a vehicle. These 66 fields were reasonably evenly distributed within the 30-km2 study area, and several fields were observable at once from discrete observation points, thereby reducing the likelihood that observers caused birds to move between study fields and so to be double-counted. Each field was visited once or twice a week on a weekday, from 24 November 1992 until 23 April 1993. At each observation point, fields were categorized according to whether they contained an unharvested crop, waste from a previous crop, had been ploughed, drilled with a winter or spring crop, or remained fallow. All fields included in the survey were potential sources of food for wood-pigeons. Newly drilled cereal fields are thought to be particularly attractive to wood-pigeons (Murton, Isaacson & Westwood 1963) so cereal fields were divided into those drilled within 1 week, and those drilled more than 1 week previous to the transect date. Any human disturbance or farming operations were recorded, and the number of wood-pigeons on the field noted. Wood-pigeon feeding behaviour varies diurnally, therefore the time of day at which each field was visited was randomized. This was done by dividing the fields into four groups and randomly choosing (i) the group observed first, and (ii) the direction (clockwise or anticlockwise) in which the route was followed.

In the subsequent three winters, the route, fields observed and frequency of observation (once or twice a week) were varied to accommodate changes in land ownership, the wishes of the farmers and the resources required for field sampling (see following section). Seventy-five fields were observed between 2 November 1993 and 29 March 1994; 64 fields were observed between 25 October 1994 and 10 April 1995; and 84 fields were observed between 16 October 1995 and 27 March 1996. Wood-pigeons may use agricultural fields as day roosts, as well as food sources, so in 1994–95 and 1995–96 an estimate was made of the proportion of each flock that was feeding. This was done by selecting 50 birds systematically from the flock (e.g. every third bird in a flock of 150), observing each individual in turn, and scoring the number out of 50 that were not immobile.

In 1995–96 twice-weekly transects were supplemented by daily observations of 45 additional fields drilled with winter cereals. These ‘mini-transects’ were undertaken in order to study more closely the 3-week period following drilling. From 3 October 1995 until 31 December 1996, fields in the study area being drilled with winter cereals were identified by observation and by communication with farmers. Thereafter, each field was visited daily (except on the days on which full transects were carried out) for 3 weeks and the numbers of feeding wood-pigeons estimated as before. The time of day a field was visited was varied by randomly choosing the starting point of these mini-transects. Some fields were on full transect and mini-transect routes, and these fields were therefore visited every day for 3 weeks. Fields not on the full transect route were visited five times per week for 3 weeks.


The level of significance for tests was set at 0·05. Values shown are means ± standard errors.

The counts of wood-pigeons in relation to crop were grossly non-normal, and contained a large number of zero values (fields with no wood-pigeons). A two-stage analysis was therefore adopted, as suggested by Manly, McDonald & Thomas (1993). The first step involved modelling the proportion of fields occupied by wood-pigeons, the second step involved modelling the number of wood-pigeons in occupied fields only. In order to allow for the multiple observations from each field, the method of Residual Maximum Likelihood (REML; Robinson 1987) was used in the second step to analyse log-transformed counts, rather than an ordinary regression model. This method of analysis is frequently used for unbalanced experiments in biological research (Robinson 1987) and ensures that the structure of the data (in this case multiple observations on each field) is used to obtain the most precise estimates of the parameters. For similar reasons, in the first step the proportions of observations with pigeons were analysed using binomial mixed models (Breslow & Clayton 1993) by means of the Genstat procedure GLMM (Generalized Linear Mixed Model; Welham 1993). This procedure essentially combines the method used to fit Generalized Linear Models (i.e. methods such as logistic regression) with the REML algorithm, thus extending the REML approach to binomially distributed data. Test statistics from both types of model are presented as Wald tests, which follow a χ2 distribution if there is no real difference present.

The REML and GLMM models, although they do estimate field effects, do not, in the form used here, allow for any temporal correlation between successive observations on the same field. Examination of the data suggests that the serial correlation present was too weak to have an appreciable effect on the results. The one possible exception to this is in the ‘mini-transect’ data of 1995–96, where the increased frequency of observations leads to stronger correlations between successive counts. In view of this, the method of Generalized Estimating Equation. (GEE; Diggle, Liang & Zeger 1994; Kenward & Smith 1995) was used for this data using a dependence correlation structure of order three (i.e. estimating correlations for observations up to 3 days apart and assuming they are zero thereafter).

Field sampling

Individual fields can differ in their attractiveness to wood-pigeons due to factors such as size of fields, proximity to roosting woods, farm buildings and hedgerows (Inglis, Thearle & Isaacson 1989). The area in hectares of each field was therefore measured using 1 : 25000 Ordinance Survey maps and an image analyser. Attractiveness may also vary with the density of seed left on the surface after drilling (Murton & Vizoso 1963). If seed treatments are avoided in the wild, the intensity of this avoidance response is likely to vary with the concentration of active substance on the seed (as it does in the laboratory; S.L. Fryday, unpublished data). Therefore, in the autumn of 1994, drilling operations on 14 fields in the study area were studied more closely. These fields were chosen from 10 farms to give as wide a geographical and temporal range as was possible within the study area. On the day of drilling, a sample of ≈ 50 seeds was taken from the hopper for residue analysis. Once drilled, the density of surface seed was estimated by counting seed in randomly placed 0·25-m2 quadrats. For each field, seed density was measured at 20 points randomly spaced along the diagonal axis of the main field (the central area of each field, where seed is drilled in parallel rows), and at 10 points randomly distributed along each headland (the edge of the field where the tractor turns, and where seed is drilled at right angles to the main field). The density of seed exposed on field surfaces may change over time due to the weather and removal by granivorous wildlife. The concentration of fonofos on treated seed may also degrade over time. Therefore, seed density was again measured on days 1, 3, 5, 8 and 14 after drilling and, at the same time, random samples were taken of seed that had been left exposed on the surface of fields, for residue analysis. In 1995–96, only day 1 measurements and samples were taken, to avoid interference with the subsequent daily wood-pigeon observations undertaken in that year.

Pesticide residues

Sampled seeds were placed in sealed labelled plastic bags and frozen at – 20 °C until they could be analysed. Up to 5 g of the sample was taken and the number of seeds counted. Samples were macerated for 2 min in 100 mL of acetone:hexane 50 : 50 v/v. The extract was then analysed using a Hewlett Packard 5890 II gas chromatograph equipped with a nitrogen-phosphorus detector (full details are given in Pascual & Hart 1997). Samples were analysed in triplicate, and variance between them was found to be less than 5%. Field samples of seeds were frequently contaminated by mud, and had absorbed a variable amount of water. For these reasons, concentrations of fonofos per seed were calculated and used to standardize residue measurements to those of dry clean seed, given in mg kg–1 seed.

Crop contents of shot birds

Although wood-pigeons are frequently observed feeding on drilled cereal fields it was important to confirm they were ingesting drilled seed. Birds feeding on fields could not be shot as this would have interfered with transects. However, organized shooting syndicates operated in the study area at weekends and arrangements were therefore made to obtain shot birds from them throughout the drilling seasons. It was not possible to control where and when the birds were shot, so this sample is necessarily biased and non-random.

Carcasses were labelled with location and habitat on which they were shot, as far as was known, and were frozen for later examination. Crops were removed and the contents examined for cereal seed. If cereal seeds were present, the contents were analysed for pesticide residues as above.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Selection of agricultural fields by wood-pigeons

Figure 1 shows the number of fields on the transect drilled with cereal seed in each of the four winters 1992–96. Where drilling dates were not known precisely, they were estimated to be mid-way between the two transect dates over which the change in the field was noted. Drilling can be carried out at any time of the year from September to April, but in the study area it tended to be in two distinct periods, autumn (October – December) and spring (February – March). In 1996 there were few spring drillings as good weather the previous autumn allowed most fields to be drilled before the end of the year.


Figure 1. The number of newly drilled cereal fields noted on transects carried out in the four winters between 1992 and 1996. The figure seems to suggest that there were more cereal fields in 1995–96 because it includes fields on both the full transects and the mini-transects (as described in the text), i.e. a wider area was covered. Each block represents 1 week.

Download figure to PowerPoint

The total number of wood-pigeons observed on transects in 1992–93 is shown in Fig. 2(a), broken down by crops and summed over periods of 1 month. In total, 25 425 wood-pigeon observations were made on 41 transects, giving a mean of 620 bird observations on each transect day, with a peak at the end of January. In the three subsequent winters the abundance of wood-pigeons varied, with mean numbers observed per transect of 1339, 381 and 196, respectively. In 1994–95 the peak number of birds occurred in November rather than January, and the reason for this is unknown.


Figure 2. (a) The total number of wood-pigeons observed on each habitat per month (sum of wood-pigeons counted on transects) in 1992–93. (b) The average area in hectares covered by each habitat for each month in 1992–93.

Download figure to PowerPoint

In 1995–96 the reduced numbers of wood-pigeons observed in open agricultural habitats was a reflection of an apparently nationwide trend (J. Harradine, personal communication), possibly due to an unusually abundant harvest of fruits and nuts in woods and hedges (I. Inglis, personal communication). The availability of different types of agricultural fields to wood-pigeons in 1992–93 is shown in Fig. 2(b). In the three subsequent winters the pattern was similar, though in 1994–96 no root waste was available beyond January.

The presence/absence of wood-pigeons on fields in relation to crop was analysed by means of a Generalized Linear Mixed Model and the results are given in Table 1. There was an overall significant difference (Wald test of differences, P < 0·05) between crops for three of the four autumn drilling periods (seasons 1992–93 1994–95 and 1995–96) and one of the spring drilling periods (season 1994–95). Wood-pigeons’ use of newly drilled cereal fields (‘cereal wk1’ in Table 1) was greater than their use of other crops for all three autumns in which there was an overall significant difference between crops (except for root waste in 1994–95). The level of significance of the difference between newly drilled cereal fields and individual crops differed between years. In the spring of 1994–95, newly drilled cereal fields were used more than any other crop, and the differences were significant (GLMM, P < 0·05) except for peas or beans.

Table 1.  The probability of use of fields by wood-pigeons in relation to crop, calculated from the Generalized Linear Mixed Model (GLMM). χ2 figures refer to 1 degree freedom (d.f.) test comparing each crop type with week 1 cereal. Data are from all transects in which newly drilled cereal fields were observed (shown in Fig. 1). Observations with disturbances were not included in the analysis (an observation being one field on one transect day). Cereal fields include all fields with and without fonofos
CropnMeanSEχ2P nMeanSEχ2P
  1. n = number of observations, mean = mean proportion of observations with wood-pigeons, χ2and P for individual crops refer to comparison with week 1 cereal.

  2. NS = not significant (P > 0·05); *P < 0·05; **P < 0·01; ***P < 0·005, significance of overall differences between crops is made using Wald's test.

Cereal wk1100·500·17   150·200·11  
Cereal wk2 + 890·070·039·27*** 1880·110·020·74NS
Ploughed420·140·064·13* 1560·170·030·01NS
Root crops170·00  NS 0    
Root waste750·470·060·19NS 240·250·090·30NS
Stubble/weeds1110·160·044·22* 200·050·050·73NS
Linseed0     60·330·210·05NS
Peas or beans0     150·270·120·46NS
Wald test of differences   χ2 19·6d.f. 5P ***χ2 7·69d.f. 7P NS  
Cereal wk1130·230·12  120·170·11   
Cereal wk2 + 330·300·080·19NS 1600·180·030·00NS
Ploughed520·270·060·05NS 1700·260·030·48NS
Root crops570·040·032·92NS 0    
Root waste1100·290·040·60NS 160·440·132·54NS
Stubble/weeds1860·230·030·00NS 540·240·060·16NS
Beans0     10·00  NS
Wald test of differences   χ2 9·73d.f. 6P NSχ2 7·14d.f. 5P NS  
Cereal wk1280·390·09   140·640·13  
Cereal wk2 + 650·290·060·93NS 1320·200·0410·1***
Ploughed630·170·055·77* 970·190·0411·8***
Root crops170·060·063·58NS 0    
Root waste610·590·061·52NS 0    
Stubble/weeds1290·220·043·75NS 200·150·086·15*
Peas or beans0     100·400·161·34NS
Drilled onions0     110·090·095·40*
Wald test of differences:   χ2 24·48d.f. 5P ***χ2 14·34d.f. 5P *  
Cereal wk1660·170·05   140·070·07  
Cereal wk2 + 2810·120·021·15NS 2680·110·020·20NS
Ploughed1420·140·030·46NS 2680·100·020·15NS
Root crops1530·020·0110·69*** 0    
Root waste1480·020·030·45NS 0    
Stubble/weeds6350·090·012·26NS 680·070·030·00NS
Drilled onions0     140·210·110·76NS
Wald test of differences:   χ2 20·32d.f. 5P ***χ2 1·48d.f. 4P NS  

When zero counts were excluded, there was no indication that the number of wood-pigeons on used fields is related to the type of crop being grown (Table 2).

Table 2.  Tests of significance for differences between crops in the numbers of wood-pigeons present, based on the REML analysis. Observations of fields with no wood-pigeons are excluded
Wald test
Autumn drilling periods
Spring drilling periods

Avoidance of fields drilled with pesticide-treated cereal seed

Cereal seed treatments used by farmers on transect fields are shown in Table 3. Of these, guazatine is the only seed treatment used, apart from fonofos, for which we are aware there is evidence of avoidance from tests with captive birds. Most seed drilled in the study area, whether treated with fonofos or not, was also treated with a fungicide (although a few farmers would drill their own, untreated seed). For the first three winters (1992–95), the percentage of newly drilled cereal fields on which wood-pigeons were observed was compared for fonofos-treated and untreated (with fonofos) fields, for each week after drilling (Fig. 3). Combining all three winters, flocks of wood-pigeons were less likely to be observed on fields drilled with fonofos-treated seed compared to untreated fields, in the first week after drilling (combining Fisher's exact tests for the 3 years,

  • image

 = 18·75, P < 0·01). The specific fields included in Fig. 3 differed between years due to changes in the transect route and crop rotation.

Table 3.  Cereal seed treatments used on transect fields 1992–96
Product namesTypeActive ingredientsRepellent in tests with captive birds?
Fonofos Seed TreatmentMicroencapsulated Organophosphorus InsecticideFonofosYes (S.L. Fryday, unpublished data)
Cerevax, Cerevax extraFungicideCarboxin Thiabendazole ImazalilNo (S.L. Fryday, unpublished data)
Panoctine, Rappor
Liquid Seed TreatmentFungicideGuazatineYes (Avery & Decker 1991)
Beret, Beret plus,FungicideFenpiclonilNot known
Beret Gold, Beret extra Fludioxonil Imazalil 
  FuberidazoleNot known
Raxil SFungicideTebuconazole TriazoxideNot known

Figure 3. The percentage of fonofos-treated and untreated fields on which wood-pigeons were observed, in relation to time after drilling, in the winters 1992–95. Figures next to data points give the numbers of individual fields used to calculate each percentage.

Download figure to PowerPoint

For 1994–95, the first winter in which guazatine was commonly used on transect fields, the above analysis was repeated using only the fields without guazatine-treated seeds. The results still hold; fewer fonofos-treated fields had feeding wood-pigeons than expected (Fisher's exact test, P < 0·05).

In 1995–96, 60 cereal fields were observed five to seven times per week for 3 weeks after drilling. In this year fewer wood-pigeons were observed in general, resulting in many more observations of fields without wood-pigeons. Therefore, data were grouped into 2-day periods (Fig. 4a). For the first 8 days after drilling, the percentage of fields on which feeding wood-pigeons were observed was less for fonofos-treated fields compared to those without fonofos. The difference in the trend over time between fonofos-treated and untreated fields is significant (Generalized Estimating Equations;

  • image

 = 6·57, P < 0·05). An index of avoidance was calculated for each field. This was the proportion of days on which wood-pigeons were observed feeding on the field for an 8-day period after drilling. Figure 5(a) shows that the degree to which fields were avoided by wood-pigeons was related to the concentration of fonofos on surface seed on the day after drilling (logistic regression; χ2 = 4·42, d.f. = 1, P < 0·05). More specifically, wood-pigeons were not observed on fields where exposed seeds had a fonofos concentration of over 150 mg kg–1 (Fig. 5a). In contrast, neither surface seed density (Fig. 5b) nor the presence of guazatine made any improvement to the model (logistic regression; χ2 = 0·06 and χ2 = 0·64, respectively, d.f. = 1, P > 0·10).


Figure 4. The use of fonofos-treated and untreated fields by wood-pigeons in relation to time after drilling in the winter of 1995–96: (a) the percentage of cereal fields on which feeding wood-pigeons were observed; and (b) mean number of feeding wood-pigeons on fields with wood-pigeons. Figures next to data points are the numbers of individual fields used to calculate the percentage or mean (sample size).

Download figure to PowerPoint


Figure 5. The percentage of observation days on which wood-pigeons were observed feeding, in relation to (a) the concentration of fonofos on surface seed and (b) the density of surface seed (number per square metre on a logarithmic scale) on day 1 after drilling, for each field. Wood-pigeon counts were undertaken between 11 October and 31 December 1995.

Download figure to PowerPoint

No significant differences in the number of wood-pigeons (as opposed to presence/absence) were apparent between fonofos-treated and untreated fields in any year. For example in 1995–96 (Fig. 4b), the year in which the most intensive observations were made, there was no effect of fonofos on the trend over time in flock size (regression analysis, adjusting for any effect of guazatine, F1,28 = 0·50, P > 0·05).

Pesticide residues on cereal seed collected in the field

The expected concentration of fonofos on treated wheat seed was 1080 mg kg–1, based on the rate of treatment approved in the UK (250 mL concentrate 100 kg seed–1; concentrate = 433 g fonofos L–1). The actual levels found on seed samples collected from drilling hoppers and drilled fields in autumn 1994 are shown in Table 4. Residue levels in both hopper and field samples of seed were well below the nominal concentration. In particular, seed left exposed on field surfaces was always less than half the nominal application rate. The decay in residue levels was calculated for drilling (hopper to day 1) and post-drilling (day 1 to day 14) for each main field. The greatest drop in residue levels occurred during drilling operations (mean of 64·31%, SE 4·7%), and the reason for this is unknown. Subsequent residue decay was slow, averaging 2% per day (mean of 27·35% from day 1 to day 14, SE 16·8%).

Table 4.  Concentrations of fonofos in samples of cereal seed collected from drilling hoppers and exposed on field surfaces during autumn 1994
Mean conc.
 (mg kg−1)SEsamplesNo. of conc.Max. conc.Min.
  1. Figures are mg of fonofos in the number of seeds equivalent to 1 kg of hopper seed. This calculation was necessary in order to take account of hydration of seed and adhering mud after drilling.

Hopper seed68548·312869298 
Day 124134·711448 77
Day 316023·610308 84
Day 517541·6 9481 
Day 818042·2 9450 
Day 1412325·910293 62

Density of drilled seed left exposed on field surfaces

The mean density of seed left exposed on the soil surface was greater on headlands than on main fields, varied widely between fields and is considered in detail in relation to drilling depth and the pesticide risk to wildlife in another paper (Pascual et al. 1999). The density of exposed seeds measured on 14 main fields in 1994–95 declined steadily at a rate of 2% per day from day 1 (2·45 seeds m–2[ ± 1·02]) to day 14 (1·66 seeds m–2[ ± 1·23]) after drilling (regression analysis, correcting for differences between fields, F1,50 = 8·49, P < 0·01). On headlands, exposed seed density declined steadily at a rate of 4% per day from day 1 (6·76 seeds m–2[ ± 1·32]) to day 14 (3·09 seeds m–2[ ± 1·35]) after drilling (regression analysis, correcting for differences between fields, F1,50 = 11·33, P < 0·005). There was no significant difference between the disappearance rates of fonofos-treated and untreated seeds (anova; F1,12 = 0·37, NS).

Crop contents of shot wood-pigeons

The crops were removed from 61 wood-pigeons shot between the months of October and April in 3 of the 4 years of the study. Table 5 shows the number of these birds that had been feeding on cereal seed and, of these, the number in which fonofos residues were detected. None of the birds shot in 1994–95 contained cereal seed, whereas approximately half of the birds shot in 1992–93 and 1995–96 had been feeding on cereal seed. The mean wet weight of the total crop contents for all dissected birds was 24·42 g (± 2·85). For the birds with cereal seed, the mean wet weight of seed removed from the crop was 8·9 g (± 1·86) and cereal seed comprised a mean of 46·6% (± 7·33) of the total crop contents. Of 24 birds with cereal seed in the crop, three had been feeding exclusively on cereal seed, the rest had been feeding on a mixed diet. Other crop contents included unidentified green vegetation, bean, pea and rape seeds, mushrooms, maize, rose hips, acorns, snails and unidentified berries. There was no indication that the six birds with fonofos residues in the crop had eaten less cereal than those with no residues (Mann–Whitney test, P > 0·10), i.e. that the cereal had been avoided. Residue levels found on the cereal in these birds’ crops were between 33 and 147 mg kg–1.

Table 5.  Frequencies of wood-pigeons with cereal seed in the crop, and of those, fonofos residues in the crop contents, from a sample of 61 wood-pigeons shot between the months of November and April during the winters of 1992–96. No birds were obtained from the year 1993–94
Year shotNo. of birds dissectedNo. of birds with cereal seedNo. of birds with fonofos residuesConcentration of fonofos (mg kg−1)
1992–931911633, 71, 75, 86, 143, 147
1994–9514 0 
Total61266As above


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

To our knowledge, this study is the first attempt to measure avoidance of pesticide-treated food by free-living birds, except for studies of the efficacy of chemicals intended as repellents (for example, Cummings et al. 1991). In addition, Dolbeer, Avery & Tobin (1994) review examples of avoidance of methiocarb (a broad-spectrum carbamate) based on reduced levels of damage. In studies with captive birds, avoidance is often clearly demonstrable (for example, Kononen, Hochstein & Ringer 1987; Bennett 1989; Fryday, Chandler-Morris & Hart 1998; Pascual, Fryday & Hart in press). The mechanisms for avoidance can be (i) via a noxious taste, or (ii) due to post-ingestional effects including nausea, lethargy, incapacitation and Conditioned Taste Aversion (in which animals learn to avoid similar foods in the future; Nicolaus et al. 1989). Captive feral pigeons offered fonofos-treated seed under realistic conditions initially consume the seed at normal rates with little evidence that it has a primary repellent effect (Fryday, Chandler-Morris & Hart 1998). Avoidance only occurs after the birds have eaten some of the seed. It is therefore unlikely to be a reaction to the taste of the treated seed and more likely to be caused by cholinesterase inhibition in the central nervous system (the primary mechanism of toxicity for organophosphorus compounds like fonofos).

Before assessing the evidence for and against avoidance in the field, it is useful to examine what might be expected from the results of laboratory studies (Fryday, Chandler-Morris & Hart 1998; Pascual, Fryday & Hart in press; S.L. Fryday, unpublished data) and other information. In the present study we found that seed exposed on the surface immediately after drilling had a mean fonofos concentration of 241 mg kg–1. At this level a 500-g wood-pigeon would need to ingest 28 g of seed to receive the median lethal dose of 6·65 mg (calculated from the LD50 for the feral pigeon; Schafer & Brunton 1979). In laboratory studies, the amount of treated seed birds will consume before the avoidance reaction begins to take effect decreases with increasing concentration of fonofos on the seed. In this study we found that fonofos residues on seed varied widely between fields (Table 4), suggesting that the extent of avoidance may be variable. Both feral and wood-pigeons have been involved in lethal poisoning incidents, though these are very rare. The Wildlife Incident Investigation Scheme (which monitors poisoning incidents for the UK Ministry of Agriculture, Fisheries and Food), has recorded six incidents between 1984 and 1996, each involving up to 30 wood-pigeons. Therefore, if avoidance does reduce consumption in the field, it does not always prevent individual birds consuming a lethal dose. Laboratory studies also suggest that avoidance is highly dependent on feeding rate: the faster birds feed, the more treated seed they will consume before the avoidance reaction comes into effect. This suggests that the extent of avoidance may vary depending on factors that affect feeding rate such as hunger (and hence weather, availability of alternative food, etc.), social factors (such as competition), the availability of food (exposed seed density) and the risk of predation. In this study we measured the density of exposed seeds and found it to be low (a mean of 2·45 seeds m–2 immediately after drilling) and variable both within and between fields. This suggests that feeding rates will in general be low, prolonging the time for avoidance to take effect, and decreasing the difference in wood-pigeon behaviour between treated and untreated fields. Taken together, these field and laboratory results imply that avoidance in the field will be less than complete and will vary widely in extent, depending on conditions.

In the present study, the evidence for avoidance in the field is the following: in 1992–95 fewer fields treated with fonofos were used by wood-pigeons in the first week after drilling than expected, and in 1995–96 the extent of the avoidance of treated fields was found to be related to the concentration of pesticide on the seed. Wood-pigeons were not observed on fields that had been drilled with seed with a fonofos concentration of over 150 mg kg–1. This is within the range at which an avoidance reaction can be measured in captive birds (S.L. Fryday, unpublished data). Seed treatments other than fonofos were used over the period of this study, and one of these (guazatine, a fungicide) is known to be avoided by red-winged blackbirds Agelaius phoeniceus in the laboratory (Avery & Decker 1991). In the first two winters guazatine was used rarely in the study area. In 1995–96 the observed avoidance reaction was found not to be related to the presence/absence of guazatine. The use of guazatine is not therefore confounding the above results.

Other data show no evidence of avoidance in the field. It is notable that evidence of avoidance was apparent in the presence or absence of flocks but not in the size of those flocks. This is because the wood-pigeon flocks on cereal fields in the study area are dynamic in size, with birds continually arriving and leaving. If avoidance caused birds to leave sooner on treated fields, we would expect smaller flocks at any one time. However, flocks were only observed on fields where the residues of fonofos on exposed seed were less than 150 mg kg–1. Laboratory tests show little avoidance below this level, so it is not surprising that no difference in flock size was detected. There also appeared to be no significant difference in the disappearance rate of treated and untreated exposed seed. However, the method used to measure mean seed density on a field was crude (20 sample points in each main field) and so would not necessarily detect changes in density over time. Finally, some fonofos-treated seed is eaten by wood-pigeons. Fonofos residues were found in the crops of shot wood-pigeons, and the concentrations measured are comparable to those found in fonofos poisoning incidents by the Wildlife Incident Investigation Scheme (M. Fletcher, personal communication). However, we should expect to find fonofos residues in wood-pigeon crops, given that the mechanism of avoidance is post-ingestional.

On balance, these results seem to suggest that avoidance in the field does occur; when many fields are examined over a period of years, treated fields are used by wood-pigeons less often than expected. As predicted by recent laboratory studies designed to more realistically reflect field conditions (above), avoidance in the field appears to be variable and incomplete. Avoidance in the field is not as strong as predicted from tests using feral pigeons in captivity under no severe hunger stress (Pascual & Hart 1997; Fryday, Chandler-Morris & Hart 1998), suggesting that extreme caution should always be exercised when extrapolating laboratory avoidance reactions to the field. Metabolism and toxicity of pesticides are found to vary between individuals with species, age, sex, nutritional status, disease, stress, interactions with other chemicals and food restriction (Villeneuve et al. 1978). Captive birds used in laboratory and aviary tests are generally of high nutritional status, disease-free and allowed unrestricted access to food. The laboratory avoidance reaction to fonofos in feral pigeons can be shown to break down when food is restricted (Pascual, Fryday & Hart in press), when birds are tested in larger group sizes and when feeding rates are increased (S.L. Fryday, unpublished data). Similarly, wild wood-pigeons may be less likely to avoid pesticide-treated cereal seed when food is scarce, when feeding in flocks and when severe weather conditions enhance hunger.

As well as being variable and incomplete, the avoidance reaction in this study appears to be restricted to the first week after drilling. Factors which may help explain this include residue decay, the disappearance rate of exposed seed and the relative preference for alternative foods. Fonofos residues were found to decay at a rate of 2% per day after drilling. Over the same time, exposed seed also disappeared at a rate of 2% per day. Both processes appear to be linear, but slow, and with no sharp declines between the first 2 weeks after drilling. This pattern of decay suggests that avoidance is most likely to be detected immediately after drilling, but they do not help explain why avoidance can be detected in week 1 but not week 2. The sampling methodology used to measure exposed seed density, however, was not intensive enough to measure spills or local areas of high seed density. However, de Leeuw et al. (1995) found that 40% of Dutch winter wheat fields investigated had ‘spill spots’ (spots where seeds are accidentally dropped instead of sown), and on these fields there was a mean of two spills per field. The possibility exists therefore that exposed seed availability is greatest in the week following drilling and declines sharply thereafter.

This hypothesis is consistent with data on the use of newly drilled cereal fields by wood-pigeons. Pair-wise comparisons show that in cases where significant differences exist, newly drilled cereal fields were used more than some alternative crops, in three out of the four autumn drilling periods and one of the four spring drilling periods. In the other periods differences were not significant. Other studies have also found that wood-pigeons prefer newly drilled cereal fields. For instance, cereal sowings were found to be preferred to cereal stubble, oilseed rape, peas and pasture during autumn and spring drilling periods between 1961 and 1985 in a study area 60 km away from ours (Murton, Isaacson & Westwood 1963; Inglis et al. 1990). The avoidance reaction may therefore be confined to a 1-week period after drilling because after this, foods other than cereal sowings are preferred. Wilson, Taylor & Muirhead (1996), in contrast, found that conventionally grown winter cereals were avoided by wood-pigeons, but organically grown winter cereals were not. However, in their study newly sown cereal fields were not distinguished from those that had germinated, and this may account for the apparent contradiction with the results of our study. Time to germination was variable in our study (between 1 and 3 weeks after drilling, depending on weather conditions; personal observation), so germination is unlikely to explain why avoidance was confined to a 1-week period after drilling.

In summary, low densities of exposed seed combined with lower than expected residue levels may explain why fonofos-treated fields are not always avoided, and localized areas of high availability of exposed seeds and a preference for newly drilled fields may explain why avoidance is confined to the first week after drilling. The amount of seed left exposed on the surface after drilling and the residue levels on drilled seed vary between and within fields, and there is therefore potential for reducing risk by adopting better risk management techniques, such as drilling seed deeper (Pascual et al. 1999). For at least some pesticides, residue levels on seed available to wild birds may consistently be well below those predicted (either due to a decline during drilling operations, as in this study, or to residue decay), and this may be an area in which existing risk assessments can be refined. Also, avoidance probably occurs in the wild, and thus evidence of repellency could potentially be used to refine risk assessments. However, we believe there is little potential to develop field tests as a routine method for confirming avoidance, because of the difficulties encountered in this study. Instead, more realistic laboratory or semi-field cage or pen trials are likely to be more useful (OECD 1996a). Finally, avoidance in the wild is incomplete, and likely to break down under conditions that will further increase natural mortality rates (stress due to poor weather, food shortage, etc.). The frequency of such conditions should therefore be taken into account in risk assessment.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study was funded by the Pesticides Safety Directorate, Ministry of Agriculture, Fisheries and Food, UK. Juan Pascual was supported by a European Union (AIR Programme) Postdoctoral Fellowship. We are grateful for access to roost woods and other study sites provided by English Nature, the warden and staff of Holme Fen and Woodwalton Fen Nature Reserves, landowners and farmers who co-operated with the study and local shooters who provided us with carcasses. Thanks also to Arnold Cooke, Ian Inglis, Lucy Rogers, Steve Fryday and Mark Clook for advice and information, and to George Watola, Adam Wilson, Denbeigh Vaughan, Justin Hart, Steven Cooke, Elizabeth Cooke and Marti Spittle for help with fieldwork.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Avery, A.L. & Decker, D.G. 1991 Repellency of fungicidal rice seed treatments to red-winged blackbirds. Journal of Wildlife Management, 55, 327334.
  • Bennett, R.S. 1989 Factors influencing discrimination between insecticide-treated and untreated foods by northern bobwhite. Archives of Environmental Contamination and Toxicology, 18, 697705.
  • Bowles, R.G. & Webster, J.P.G. 1995 Some problems associated with the analysis of the costs and benefits of pesticides. Crop Protection, 14, 593600.
  • Breslow, N.E. & Clayton, D.G. 1993 Approximate inference in generalized linear mixed models. Journal of American Statistical Association, 88, 925.
  • Caldwell, T.H. & Richardson, S.J. 1975 Field behaviour of lowland peats and organic soils. Soil Physical Conditions and Crop Production. Ministry of Agriculture, Fisheries and Food Technical Bulletin 29, pp. 94111. HMSO, London.
  • Campbell, L.H., Avery, M.I., Donald, P., Evans, R.E., Green, R.E. & Wilson, J.D. 1997A Review of the Indirect Effects of Pesticides on Birds. Report to the Department of the Environment, the Joint Nature Conservation Committee and English Nature prepared by the Royal Society for the Protection of Birds. JNCC, Peterborough.
  • Cooke, A.S. 1988 Poisoning of woodpigeons on Woodwalton Fen. Field Methods for the Study of Environmental Effects of Pesticides (eds M.P.Greaves, B.D.Smith & P.W.Greig-Smith), pp. 297301. British Crop Protection Council, Thornton Heath.
  • Cummings, J.L., Mason, J.R., Otis, D.L. & Heisterberg, J.F. 1991 Evaluation of dimethyl and methyl anthranilate as a Canada goose repellent on grass. Wildlife Society Bulletin, 19, 184190.
  • Dewar, A.M. & Asher, M.J.C. 1994 A European perspective on pesticide seed treatments in sugar beet. Pesticide Outlook, 5, 1117.
  • Diggle, P.J., Liang, K.-Y. & Zeger, S.L. 1994Analysis of Longitudinal Data. Oxford University Press, Oxford.
  • Dolbeer, R.A., Avery, M.L. & Tobin, M.E. 1994 Assessment of field hazards to birds from methiocarb applications to fruit crops. Pesticide Science, 40, 147161.
  • Donald, P.F. & Forrest, C. 1995 The effects of agricultural change on population size of corn buntings Miliaria calandra on individual farms. Bird Study, 42, 205215.
  • Ekstrom, G. & Bernson, V. 1995 Swedish pesticide policies 1972–93: risk reduction and environmental charges. Reviews of Environmental Contamination and Toxicology, 141, 2770.
  • Evans, J., Wilson, J.D. & Browne, S.J. 1994 Habitat selection and breeding success of skylark Alauda arvensis on organic and conventional farmland. British Trust for Ornithology Research Report 154, Part III. British Trust for Ornithology, Thetford.
  • Feare, C.J. 1989 The changing fortunes of an agricultural bird pest: the European starling. Agricultural Zoology Reviews, 3, 317342.
  • Fletcher, M.R. & Hunter, K. 1992Pesticide Poisoning of Animals 1992: Investigations of Suspected Incidents in the United Kingdom. MAFF Publications, London.
  • Fletcher, M.R., Hunter, K. & Barnett, E.A. 1992Pesticide Poisoning of Animals 1993: Investigations of Suspected Incidents in the United Kingdom. MAFF Publications, London.
  • Fletcher, M.R., Hunter, K. & Barnett, E.A. 1995Pesticide Poisoning of Animals 1994: Investigations of Suspected Incidents in the United Kingdom. MAFF Publications, London.
  • Fletcher, M.R., Hunter, K., Barnett, E.A. & Sharp, E.A. 1996Pesticide Poisoning of Animals 1995: Investigations of Suspected Incidents in the United Kingdom. MAFF Publications, London.
  • Fletcher, M.R., Hunter, K., Quick, M.P., Thompson, H.M. & Greig-Smith, P.W. 1991Pesticide Poisoning of Animals 1990: Investigations of Suspected Incidents in Great Britain. MAFF Publications, London.
  • Fryday, S.L., Chandler-Morris, S.A. & Hart, A.D.M. 1998 Presentation method and the avoidance of fonofos-treated seed by captive birds. Bulletin of Environmental Contamination and Toxicology, 61, 448454.
  • Fuller, R.J., Gregory, R.D., Gibbons, D.W., Marchant, J.H., Wilson, J.D., Baillie, S.R. & Carter, N. 1995 Population declines and range contractions among lowland farmland birds in Britain. Conservation Biology, 9, 14251441.
  • Greig-Smith, P.W. 1987 Hazards to wildlife from pesticide seed treatments. Application to Seeds and Soil (ed. T.Martin), pp. 127134. British Crop Protection Council, Thornton Heath.
  • Inglis, I.R., Isaacson, A.J., Thearle, R.J.P. & Westwood, N.J. 1990 The effects of changing agricultural practice upon woodpigeon Columba palumbus numbers. Ibis, 132, 262272.
  • Inglis, I.R., Thearle, R.J.P. & Isaacson, A.J. 1989 Woodpigeon (Columba palumbus) damage to oilseed rape. Crop Protection, 8, 299309.
  • Kenward, M.G. & Smith, D.M. 1995 Computing the generalized estimating equations with quadratic covariance estimation for repeated measures. Genstat Newsletter, 32, 5062.
  • Kononen, D.W., Hochstein, J.R. & Ringer, R.K. 1987 Avoidance behaviour of mallards and northern bobwhite exposed to carbofuran-contaminated food and water. Environmental Toxicology and Chemistry, 6, 4150.
  • De Leeuw, J., Gorree, M., De Snoo, G.R., Tamis, W.L.M., Van Der Poll, R.J. & Luttik, R. 1995 Risks of granules and treated seeds to birds on arable fields. Centre of Environmental Science report, 118, Leiden University, The Netherlands.
  • Levitan, L., Merwin, I. & Kovach, J. 1995 Assessing the relative environmental impacts of agricultural pesticides: the quest for a holistic method. Agriculture, Ecosystems and Environment, 55, 153168.
  • Manly, B., McDonald, L. & Thomas, D. 1993Resource Selection by Animals. Statistical Design and Analysis for Field Studies. Chapman & Hall, London.
  • Mason, H.J. 1975An Introduction to the Black Fens. H.J. Mason, Cambridge.
  • Murton, R.K., Isaacson, A.J. & Westwood, N.J. 1963 The feeding ecology of the wood-pigeon. British Birds, 56, 345375.
  • Murton, R.K. & Vizoso, M. 1963 Dressed cereal seed as a hazard to wood-pigeons. Annals of Applied Biology, 52, 503517.
  • Nicolaus, L.K., Herrera, J., Nicolaus, J.C. & Dimmick, C.R. 1989 Carbachol as a conditioned taste aversion agent to control avian depredation. Agriculture, Ecosystems and Environment, 26, 1321.
  • OECD 1996a Report of the SETAC/OECD Workshop on Avian Toxicity Testing. held in Pensacola, Florida, USA, in December, 1994. Report OCDE/GD (96), 166, Series on Testing and Assessment no. 5, OECD, Paris.
  • OECD 1996b Testing for avoidance. Report of the SETAC/OECD Workshop on Avian Toxicity Testing. held in Pensacola, Florida, USA, in December, 1994. Report OCDE/GD (96), 166, Series on Testing and Assessment no. 5., pp. 6396. OECD, Paris.
  • Pascual, J.A., Fryday, S.L. & Hart, A.D.M. in press Effects of food restriction on food avoidance and risk of acute poisoning of captive feral pigeons from fonofos-treated seeds. Archives of Environmental Contamination and Toxicology, in press.
  • Pascual, J.A. & Hart, A.D.M. 1997 Exposure of captive feral pigeons to fonofos-treated seed in a semifield experiment. Environmental Toxicology and Chemistry, 16, 25432549.
  • Pascual, J.A., Hart, A.D.M., Saunders, P.J., McKay, H.V., Kilpatrick, J. & Prosser, P. 1999 Agricultural methods to reduce the risk to birds from cereal seed treatments. I. Sowing depth manipulation. Agriculture, Ecosystems and Environment, 72, 5973.
  • Petersen, B. 1994 Interaction between birds and agriculture in Denmark: from simple counts to detailed studies of breeding success and foraging behaviour. Bird Numbers 1992. Distribution, Monitoring and Ecological Aspects. Proceedings of the 12th International Conference of IBCC and EOAC, Noordwijkerhout, the Netherlands (eds E.J.M.Hagemeijer & T.J.Verstrael), pp. 4956. Statistics Netherlands, Voorburg/Heerlen & SOVON, Beek-Ubbergen.
  • Robinson, D.L. 1987 Estimation and use of variance components. Statistician, 36, 314.
  • Saris, F.A., Van Dijk, A., Hustings, F., Lensink, R. & Van Scarenburg, K. 1994 Breeding birds in the changing agricultural environment in the Netherlands. Bird Numbers 1992. Distribution, Monitoring and Ecological Aspects. Proceedings of the 12th International Conference of IBCC and EOAC, Noordwijkerhout, the Netherlands (eds E.J.M.Hagemeijer & T.J.Verstrael), pp. 7585. Statistics Netherlands, Voorburg/Heerlen & SOVON, Beek-Ubbergen.
  • Schafer, E.W. Jr & Brunton, R.B. 1979 Indicator bird species for toxicity determinations: is the technique usable in test method development? Vertebrate Pest Control and Management Materials (ed. J.R.Beck), pp. 157168. STP 680, American Society for Testing and Materials, Philadelphia.
  • Stanley, P.I. & Bunyan, P.J. 1979 Hazards to wintering geese and other wildlife from the use of dieldrin, chlorfenvinphos and carbophenothion as wheat seed treatments. Proceedings of the Royal Society of London B, 205, 3145.
  • Tucker, G.M. 1992 Effects of agricultural practices on field use by invertebrate-feeding birds in winter. Journal of Applied Ecology, 29, 779790.
  • Villeneuve, D.C., Van Logten, M.J., Den Tonkelaar, E.M., Rauws, A.G., Kroes, R. & Van Esch, G.J. 1978 The combined effect of food restriction and parathion exposure in rats. Archives of Environmental Contamination and Toxicology, 7, 3745.
  • Welham, S. 1993Procedure GLMM. Genstat 5 Procedure Library Manual, NAG, Oxford.
  • Wilson, J.D., Taylor, R. & Muirhead, L.B. 1996 Field use by farmland birds in winter: and analysis of field type preferences using resampling methods. Bird Study, 43, 320332.

Received 14 March 1998; revision received 4 February 1999