Prey abundance and the strength of interference in a foraging shorebird

Authors

  • Patrick Triplet,

    1. RN Baie de Somme, SMACOPI, 1 Place Amiral Courbet, 80100 Abbeville, France; and
    Search for more papers by this author
  • Richard A. Stillman,

    Corresponding author
    1. Institute of Terrestrial Ecology, Furzebrook Research Station, Wareham, Dorset BH20 5AS, UK
      ‡ Correspondence author: Dr R.A. Stillman, Institute of Terrestrial Ecology, Furzebrook Research Station, Wareham, Dorset BH20 5AS, UK. Tel: 01929 551518. E-mail: r.stillman@ite.ac.uk
    Search for more papers by this author
  • John D. Goss-Custard

    1. Institute of Terrestrial Ecology, Furzebrook Research Station, Wareham, Dorset BH20 5AS, UK
    Search for more papers by this author

‡ Correspondence author: Dr R.A. Stillman, Institute of Terrestrial Ecology, Furzebrook Research Station, Wareham, Dorset BH20 5AS, UK. Tel: 01929 551518. E-mail: r.stillman@ite.ac.uk

Abstract

Interference is an important component of food competition but is often difficult to detect and measure in natural animal populations. Although interference has been shown to occur between oystercatchers Haematopus ostralegus L. feeding on mussels Mytilus edulis L., four previous studies have not detected interference between oystercatchers feeding on cockles Cerastoderma edule L. In contrast, this study detected interference between cockle-feeding oystercatchers in the Baie de Somme, France. Prey stealing (kleptoparasitism), one of the main causes of interference between mussel-feeders, also occurred between oystercatchers in the Baie de Somme. The kleptoparasitism rate was related to the natural variation in the food supply, tending to be higher when cockles were rare. Feeding rate was negatively related to competitor density, so providing evidence for interference, but, as in mussel-feeders, only above a threshold density of about 50–100 birds ha−1. The strength of interference at a fixed competitor density was related to the cockle food supply, usually being greater when cockles were rare. Previous studies probably failed to detect interference between cockle-feeders because competitor densities were too low, or cockles were too abundant, or because they were not conducted during late winter when interference is most intense. The study shows that natural variation in the food supply can influence the strength of interference within an animal population and provides support for those behaviour-based interference models which predict that the strength of interference will be greatest when competitor densities are high and prey scarce.

Introduction

Interference, the short-term, reversible decline in intake rate due to the presence of competitors (Goss-Custard 1980; Sutherland 1983), is an important component of intraspecific food competition and one of the key factors thought to determine the distribution of foraging animals (e.g. Sutherland 1983; Parker & Sutherland 1986; Korona 1989; Holmgren 1995; Moody & Houston 1995). Recently, theoretical models based on forager behaviour and possible mechanisms of interference have predicted that the strength of interference (defined as the proportional change in intake rate caused by a proportional change in density) should vary with competitor density, being insignificant at low densities but steadily increasing in intensity as density rises (Ruxton, Gurney & de Roos 1992; Moody & Houston 1995; Stillman, Goss-Custard & Caldow 1997). Behaviour-based interference models also predict that interference will be more intense when prey are rare. This occurs either because predators spend more time searching for prey when prey are scarce and while in this state may interfere with one another (Ruxton et al. 1992; Moody & Houston 1995; Moody & Ruxton 1996), and/or because prey stealing (and hence interference) is only profitable when prey are scarce (Stillman et al. 1997). Two recent experimental field studies on snow buntings Plectrophenox nivalis L. (Dolman 1995) and blackbirds Turdus merula L. (Cresswell 1998) in which food abundance was varied artificially also showed that interference was more intense at low to mid prey densities. These studies support the predictions of behaviour-based models but do not show whether the natural variation in food abundance is sufficient to cause changes in the strength of interference.

To date, most studies of interference in birds have been conducted on oystercatchers Haematopus ostralegus L. feeding on mussels Mytilus edulis L. In this system the degree of interference has been related to a wide range of factors including the dominance (e.g. Ens & Goss-Custard 1984; Goss-Custard et al. 1988), feeding method and age of individuals, stage of the season and stage of the tidal exposure period (Goss-Custard et al. 1987; Stillman et al. 1996). Recently, the strength of interference between mussel-feeders has been shown to be greater at higher competitor densities (Stillman et al. 1996), as predicted by the theoretical behaviour-based models of interference. These field studies leave no doubt that interference occurs between oystercatchers feeding on mussels. In contrast, four studies of oystercatchers feeding on another intertidal bivalve, the cockle Cerastoderma edule L. (Goss-Custard 1977; Sutherland & Koene 1982; Ens et al. 1996; Norris & Johnstone 1998) have found little or no evidence for interference. Yet, as many cockle feeding birds forage where cockles, and so competitors, are scarce, the distribution of oystercatchers on their cockle feeding grounds seems to imply that interference does occur (Goss-Custard, West & Sutherland 1996). Additionally, the combination of handling time and prey encounter rate in cockle-feeding oystercatchers suggests that prey stealing, one of the major components of interference between mussel-feeders, should also occur between cockle-feeders, at least when cockles are rare and large (Stillman et al. 1997).

In this paper we provide evidence that interference occurs between cockle-feeding oystercatchers under certain conditions in the Baie de Somme, France. We examine whether the natural variation in the cockle food supply is linked to the strength of interference, test the predictions of behaviour-based interference models and suggest reasons why interference occurs between cockle-feeders in the Baie de Somme but has not been detected previously.

Methods

The field data were collected between October 1988 and March 1997 in the Baie de Somme, north-west France (50·14 °N/1·33 °E), an estuary with a total area of 7000 ha, including over 4000 ha of mud and sandflats. The study site was located in a nature reserve in the north-west of the estuary, in an area of muddy-sand where cockles dominated other bivalves in terms of both density and biomass. Inside the reserve birds are protected, but outside hunting occurs on a range of species, including oystercatchers. As a consequence, virtually no oystercatchers occur outside the reserve even though suitable habitat exists, whilst within the reserve they occur at very high densities (usually between 80 and 250 birds ha−1, but sometimes exceeding 500 birds ha−1). See Triplet 1994a for further details of the study area.

Oystercatcher feeding rates were recorded between October and March during five winter periods – 1988/89 1989/90 1994/95 1995/96 1996/97 – from birds foraging in 25 × 25 m square quadrats. Observations were made by selecting a focal bird, identifying it as either an immature (2nd to 4th winter) or adult, and recording its feeding rate in one of two ways: in 1988/89 and 1989/90, the time taken to consume three cockles was measured and then converted to the number of cockles consumed per 5 min; in the subsequent years, feeding rate was measured directly as the number of cockles consumed during a 5-min observation period. There was no evidence that these contrasting methods of measuring feeding rate influenced the patterns found in the different years. Intake rate (i.e. the rate of consuming prey flesh) could not be estimated as many cockles were opened within the substrate so making it impossible to estimate their size accurately. Handling time was measured occasionally in each year, except 1994/95 in which it was not measured at all, as the time taken to open and consume a cockle. In each year except 1994/95, the number of kleptoparasitic disputes between the focal bird and other oystercatchers was also recorded during each observation period. The competitor density experienced by a focal bird was recorded from the number of other oystercatchers within the quadrat in which the bird was foraging. Oystercatcher densities varied widely within individual quadrats, generally being higher on the advancing and receding tides than at low water, but also frequently reaching high levels at low tide due to influxes of birds from disturbed areas. Additionally, in 1996/97 oystercatcher densities increased in late winter due to a large influx of birds displaced from the Netherlands by a prolonged spell of cold weather (P. Triplet, personal observation). Gulls frequently steal cockles from oystercatchers in the Baie de Somme (Triplet 1994b) and in 1996/97 particularly large numbers of gulls were present (P. Triplet, personal observation). To minimize the potentially confounding effect of gulls on interference between oystercatchers, all foraging observations in which the focal bird was attacked by a gull or in which gulls were present in a quadrat were excluded from the analysis. The density and mean size of cockles in the substrate were recorded during November in each year. Additionally, cockle density was recorded during each month of winter during 1994/95 1995/96 and 1996/97. Stage of the season (days since 1 October) and stage of the tidal cycle (minutes since previous high water) were also recorded for each observation. Table 1 lists summary statistics for the variables used in the analysis.

Table 1.  Summary statistics for variables used in the analysis
VariablenMeanSDRange
Annual kleptoparasitism rate (cockle consumed−1)40·0410·0300·01–0·08
Probability of kleptoparasitism (300 s−1)23340·0650·2460 or 1
Feeding rate (cockles 300 s−1)26975·75·20–60
Competitor density (ha−1)2697164·4224·70–2176
November cockle density (m−2)51186·2933·9329–2498
November mean cockle size (mm)522·85·316–28
Monthly cockle density (m−2)17691196·21255·90–4470
Age of focal bird (1 = adult; 0 = immature)26970·530·50 or 1
Stage of the season (days since 1 October)269775·942·35–145
Stage of the tidal cycle (min)2697351·5156·230–720
Handling time (s)144814·29·22–84

Results

Variation in the cockle population

The cockle population in the study area at the start of winter varied widely between the different study years (Fig. 1). In November 1989, 1990 and 1996 populations were comprised of large individuals but the overall density was higher in 1989 than 1990 and lowest in 1996. In November 1994 and 1995 most cockles were small, but of low density in the first winter and high density in the second. A wide range of feeding conditions were therefore available for oystercatchers throughout the study.

Figure 1.

The size distribution and abundance of cockles in the study area during November at the start of each winter.

Kleptoparasitism and the cockle population

Interference in mussel-feeding oystercatchers occurs because subdominant individuals lose an increasing number of prey at higher competitor densities due to kleptoparasitic attacks from other oystercatchers, and also find mussels at a lower rate, possibly due to increased avoidance of potential aggressors (Ens & Cayford 1996). Kleptoparasitism also occurred between cockle-feeders throughout the present study; 209 kleptoparasitic attacks were observed at an overall rate of 0·032 attacks per cockle consumed. The attack rate varied between the four years in which it was measured and was significantly (linear regression; n = 4 years; P < 0·05) negatively related to the abundance of the cockle food supply at the start of winter (Fig. 2a). In contrast, annual attack rate was not related to the mean size of cockles in November, mean oystercatcher competitor density or mean feeding rate, when either modelled individually or in combination with cockle abundance (linear regression; n = 4 years; P > 0·1 for all tests).

The relationship between the kleptoparasitism rate and the cockle population was further investigated within individual years. Logistic regression was used to relate the presence or absence of kleptoparasitism during single foraging observations to feeding rate, bird age, stage of the season, stage of the tidal cycle, competitor density and cockle density. Feeding rate was included in the analysis to account for the possibility that kleptoparasitism was more likely within foraging observations in which more cockles were consumed. Competitor density had a high proportion of low values and so was log10 transformed to normalize its distribution. Both linear and squared terms were used for stage of the season to account for possible nonlinear effects of this variable. The analysis was restricted to 1995/96 and 1996/97 because in these years both kleptoparasitism and the cockle population were recorded throughout winter; in the other years only one or neither of the variables were recorded. The two years used in the analysis were extremes both in terms of kleptoparasitism rate and cockle density; in 1995/96 kleptoparasitism was rare and cockles frequent where in 1996/97 kleptoparasitism was frequent and cockles rare (Fig. 2a). Within 1995/96 the probability of kleptoparasitism occurring was unrelated to cockle density but was related to stage of the season (Table 2). In contrast, in 1996/97 the presence of kleptoparasitism was not seasonally related but was negatively related to cockle abundance (Table 2, Fig. 2b). Therefore, the negative relationship between kleptoparasitism and cockle abundance found between winters was also found within 1996/97 but not within 1995/96. A possible explanation for the lack of relationship within 1995/96 is that in this year the kleptoparasitism rate was extremely low, kleptoparasitism occurred in only 27 of 1087 foraging observations, and this may have reduced the chance of detecting any relationship.

Table 2.  Factors associated with the probability of kleptoparasitism occurring between cockle-feeding oystercatchers. The analysis was restricted to 1995/96 and 1996/97 because only in these years were both the presence of kleptoparasitism and the cockle food supply recorded throughout winter. The values show coefficients estimated from a logistic regression analysis of the probability of an attack occurring during a 300 s foraging observation against all explanatory variables
 1995/96
n = 1087
1996/97
n = 319
  • **

    P < 0·01

Intercept−9·47**−2·33
Feeding rate 0·0139−0·0030
Age of focal bird−0·187−0·212
Days since 1 October 0·151** 0·014
(Days since 1 October)2−0·00107**−0·00012
Time after high water−0·00038 0·00082
log10 (Competitor density +1) 0·485 0·372
Monthly cockle density 0·00021−0·00255**

Interference

The presence of kleptoparasitism shows that one potential mechanism of interference is present in the oystercatcher–cockle system. Additionally, the relationships between kleptoparasitism rate and the cockle food supply suggest that interference may be more severe when cockles are rare. However, the presence of kleptoparasitism does not necessarily mean that feeding rates will be influenced by competitor density (i.e. that interference occurs in the system). By definition interference can only be measured as the change in feeding rate caused by a change in competitor density.

Feeding rate may be influenced by a number of variables other than competitor density. In order to take account of such factors, and to reduce the possibility of generating a spurious relationship between feeding rate and competitor density, bird age, stage of the season and the tidal cycle and cockle density and size were also included in the analysis. Some variables were included as both untransformed and squared values in order to account for possible nonlinear relationships. Interference is usually measured from the relationship between log feeding (or intake) rate and log competitor density. Therefore, for consistency with previous studies and to normalize the distribution of these variables, both were log10 transformed. The detection of interference between mussel-feeding oystercatchers requires large sample sizes because feeding rates vary widely over the relatively short duration of foraging observations. Cockle feeding rates also varied widely (Table 1) and so data from all 5 years were combined to increase the chance of detecting interference.

The effect of all variables, except competitor density, on feeding rate was initially modelled in order to determine whether these variables could potentially generate a spurious interference relationship. All variables were significantly (P < 0·001) related to feeding rate (Table 3a). Feeding rate was higher in adults than in immatures, increased during autumn before decreasing later in winter (more than 80 days after 1 October), decreased with time through the tidal cycle, increased at a decelerating rate with increased November cockle density and decreased with increased cockle size. As all variables were related to feeding rate all were included in subsequent analyses.

The addition of competitor density to the model had no significant effect on the coefficients of the other variables (Table 3b). Feeding rate was nonlinearly related to oystercatcher competitor density (both linear and squared terms significant (P < 0·001) in Table 3b) and reached a maximum at a density of about 50 birds ha−1 (maximum point in quadratic relationship). Feeding rates at competitor densities less than 50 birds ha−1 were significantly lower than those at densities in the range 50–100 birds ha−1 (anova of log10 (feeding rate +1); F = 19·6; d.f. = 1,1498; P < 0·001; Fig. 3a), while above 50 birds ha−1, feeding rates declined with increased competitor density (Fig. 3a). Feeding rates were probably reduced in birds at low competitor densities because of increased levels of vigilance compared to birds at high densities. The point at which feeding rate started to decline with increased density is within the range of 50–150 birds ha−1 observed in mussel-feeding oystercatchers (Stillman et al. 1996). In order to remove the potentially confounding effect of increased vigilance at low competitor densities and to search for interference in the density range in which it is found in mussel-feeders, the analysis was repeated with the subset of data collected at competitor densities over 50 birds ha−1. Across this density range, feeding rate was linearly and negatively related to competitor density both when regressed singly (Fig. 3b) and when related to competitor density in combination with all other variables (Table 3c). Evidence for interference was therefore found but only at competitor densities over 50 birds ha−1.

Interference and the cockle population

As the kleptoparasitism rate was negatively related to cockle density, interference was expected to be stronger when cockles were rare. Preliminary evidence for such a relationship was sought by repeating the previous analysis (Table 3c) separately for each of the five study years. The cockle population varied widely between the different study years and so it was anticipated that the strength of interference would also vary between years. November cockle density and size were not included as explanatory variables because they did not vary within single years. The previous analysis showed that all variables were significantly related to feeding rate and so all were included in each within year analysis even if they were not significantly related to feeding rate. In general, the within year analyses showed the same results as the between year analysis although not all relationships were significant (Table 4). Adults had significantly higher feeding rates than immatures in 3 years, feeding rate was related to the stage of the season in 3 years and feeding rate decreased through the tidal cycle in 2 years. Evidence for interference, a negative relationship between feeding rate and competitor density, was found in 1989/90 1990/91 and 1994/95, whereas in 1995/96 feeding rate was not related to competitor density and in 1996/97 feeding rate increased with increased competitor density. Therefore, as expected, the strength of interference did vary between the study years. However, the strength of interference within in each year was not related to the cockle food supply measured as either November cockle density (Y = −0·138 [se = 0·178] + 0·000034 [se = 0·000123]X;P > 0·5) or size (Y = −0·049 [se = 0·513] − 0·00217 [se = 0·02199]x;P > 0·5). The lack of relationship between interference and cockle density was due to the apparent absence of interference in 1996/97 in which feeding rate was actually positively related to competitor density (Table 4). This result was surprising because in this year the kleptoparasitism rate was higher than in any other (Fig. 2a) implying that oystercatchers were in fact interfering with one another.

Table 4.  Factors associated with the feeding rate (log10 (cockles consumed 300 s−1 + 1)) of cockle-feeding oystercatchers in each of the five study years. The values show estimated coefficients for each variable in each year. Models were based on observations collected at competitor densities greater than 50 birds ha−1
 1989/90
n = 402
1990/91
n = 206
1994/95
n = 286
1995/96
n = 866
1996/97
n = 126
  • *

    P < 0·05

  • **

    P < 0·01.

Intercept 1·33** 1·36** 0·51** 0·274**−0·109
Age of focal bird 0·075**−0·002 0·245** 0·074* 0·018
Days since 1 October−0·0058**−0·0010 0·0082** 0·0149** 0·0031
(Days since 1 October)2 0·0000323** 0·0000113−0·0000454**−0·0000931**−0·0000243
Time after high water−0·000050−0·000208**−0·000031−0·000423** 0·000092
log10 (competitor density +1)−0·149**−0·287**−0·261** 0·015 0·192*

A potential problem with the previous analysis is that it divided the data into separate years and so sample sizes were small (in particular the result for 1996/97 was based on only 126 observations). Feeding rates vary widely over short foraging observations and so large sample sizes are usually required in order to detect interference and account for other factors potentially affecting feeding rate. Therefore, the interaction between the cockle food supply and interference was further investigated with the data for all years combined. The analysis in Table 3c was repeated with one additional variable which measured the combined effect of cockle density and competitor density on feeding rate. This model therefore allowed for the possibility that the strength of interference was related to the food supply. Two separate analyses were performed in which the food supply was measured as either the density of cockles present in November or the density of cockles in the quadrat in which a foraging observation was made. The first analysis (Table 3d) used data from all years but the second (Table 3e) was restricted to 1994/95 1995/96 and 1996/97 because cockle densities were only recorded throughout winter during these years. The general results of the two analyses were the same except the signs of the cockle density and cockle density squared coefficients differed (Tables 3d and e): when the food supply was measured as November cockle density, feeding rate increased at a decelerating rate with increased cockle density, but when measured as monthly cockle density, feeding rate increased at an increasing rate with increased cockle density. However, the general pattern that feeding rate increased with increased cockle abundance was the same in both analyses. The effects of all other variables were the same in both analyses and also the same as found in previous analyses (Table 3a–c). In both analyses the interaction between cockle density and competitor density was significantly (P < 0·001) positively related to feeding rate (Tables 3d and e). Feeding rate was still significantly (P < 0·001) negatively related to competitor density and so both analyses showed that increased competitor density reduced feeding rate but to a lesser extent when cockles were abundant. This result persisted when feeding rate was related to the terms containing competitor density and cockle density in the absence of other variables (Fig. 4). There was therefore evidence of an interaction between cockle abundance and the strength of interference, both when abundance was measure in November and throughout winter, in the same direction as that found between cockle abundance and kleptoparasitism rate; both the kleptoparasitism rate and strength of interference were greater when cockles were rare. This result occurred even though the data from 1996/97 did not show the expected pattern because data from this year comprised only a small fraction of the overall sample (7% and 10% of the first and second analyses, respectively). A similar analysis using an interaction term between November cockle size and competitor density failed to show a significant effect of cockle size on the strength of interference.

Figure 4.

The effect of cockle abundance on the relationship between feeding rate and competitor density in cockle-feeding oystercatchers. Cockle abundance was measured as either (a) cockle density recorded in November at the start of each winter, or (b) cockle density recorded in each month of winter during 1994/95 1995/96 and 1996/97 (monthly cockle densities were not recorded during the remaining years). Regression models were used to relate feeding rate to competitor density, cockle abundance and the interaction between competitor density and cockle abundance: (a) Y = 0·976 [se = 0·137] + 0·000470 [se = 0·000081]X1 − 0·000000228 [se = 0·00000002]X12 − 0·284 [se = 0·063]X2 + 0·000102 [se = 0·000030]X1X2; where X1 = November cockle density and X2 = log10 (competitor density +1); n = 1886; P < 0·001 for all coefficients; (b) Y = 0·931 [se = 0·093] − 0·000341 [se = 0·000061]X1 + 0·000000060 [se = 0·00000001]X12 − 0·148 [se = 0·041]X2 + 0·000093 [se = 0·000022]X1X2; where X1 = monthly cockle density and X2 = log10 (competitor density +1); n = 1278; P < 0·001 for all coefficients. For presentation the fitted relationships and observations are shown for two cockle abundance ranges, less than or greater than 1000 cockles m−2 recorded either (a) in November at the start of winter, or (b) during any month of winter, and five competitor density ranges. The symbols show mean observed feeding rate (with associated standard errors) within each cockle abundance and competitor density range and the lines show the fitted relationships.

Feeding rate vs. intake rate

The previous results showed that a potential mechanism of interference occurred between cockle-feeding oystercatchers and that feeding rate decreased with increased competitor density above a threshold value which was close to that found in mussel-feeders. These factors in combination suggest that interference occurred in the system. A potential problem, however, is that feeding rate rather than intake rate was recorded because the size of cockles consumed could not always be estimated. If larger cockles – which take longer to handle (Triplet 1994a) – were consumed where competitor densities were higher, a spurious interference relationship might be generated: feeding rate would have decreased because of the longer handling times of the larger cockles rather than because of interference. The effect of the size of cockles in the substrate was taken into account in the analysis, so reducing the chance of a spurious relationship, but as a further test the relationship between handling time and competitor density was studied. Handling time was not measured consistently throughout the study and so was not included in the main analysis. On the occasions on which it was measured, handling time was not positively, but negatively related to competitor density (linear regression; handling time = 18·5 (se = 0·7) − 2·26 (se = 0·36) log10 (competitor density +1); n = 1448; P < 0·01). The most probable effect of this would have been to increase, rather than to decrease, feeding rates at higher competitor densities, i.e. to generate the opposite relationship to that due to interference. Therefore, if intake rate instead of feeding rate had been recorded, the estimated strength of interference may have been greater.

Discussion

The results provide a range of evidence that interference occurs between oystercatchers feeding on cockles in the Baie de Somme. Kleptoparasitism, one of the main causes of interference between mussel-feeding oystercatchers (Ens & Cayford 1996), was observed during the study. Feeding rates declined at high competitor densities even when the possible confounding effects of other variables related to feeding rate were removed. The shape of the interference function was similar to that previously found in mussel-feeding oystercatchers (Stillman et al. 1996). Finally, both the strength of interference and the kleptoparasitism rate usually increased when the cockle food supply was rare, as predicted by behaviour-based interference models (Ruxton et al. 1992; Moody & Houston 1995; Stillman et al. 1997).

Although most analyses showed that the kleptoparasitism rate and strength of interference increased as cockle density decreased, two results did not fit the general pattern. In 1995/96 the kleptoparasitism rate was not related to the food supply and in 1996/97 feeding rate increased, rather than decreased, with competitor density even though cockles were rare and kleptoparasitism frequent. The result from 1995/96 can probably be explained by the overall frequency of kleptoparasitism in this year; virtually no kleptoparasitic attacks occurred, as expected from the abundant cockle population. This may have reduced the chance of finding an association with the cockle food supply. The result from 1996/97 was surprising but might be linked to three unusual characteristics of the year. First, feeding rates across the full range of competitor density were very low (1·9 cockles 300 s−1, compared to the 4·6 cockles 300 s−1 in the next lowest year 1994/95). Detecting a further reduction in feeding rate due to interference would therefore be difficult. Secondly, gulls were more abundant than in any other year and frequently stole cockles from oystercatchers. Although observations with gulls present were excluded from the analysis, their general abundance throughout winter may have influenced oystercatcher behaviour. For example, the advantage of flocking in terms of reduced vigilance against gull attacks may have outweighed the disadvantage of increased interference. Third, oystercatcher densities increased in late winter due to an influx of birds displaced from further north by cold weather. Birds are thought to make such movements because they are losing weight and approaching starvation (Hulscher, Exo & Clark 1996). If, in desperation, these birds could increase their feeding rate (as time-stressed cockle-feeding oystercatchers have been shown to do (Swennen, Leopold & Bruijn 1989)) then high feeding rates would be associated with high competitor densities.

Interference has frequently been detected among oystercatchers feeding on mussels (Ens & Cayford 1996) but has not previously been detected in cockle-feeding birds (Goss-Custard 1977; Sutherland & Koene 1982; Ens et al. 1996; Norris & Johnstone 1998). The lack of previous evidence of interference may, however, be explained by a combination of the results of this study, studies of interference in mussel-feeders and the predictions of behaviour-based interference models. In this study, interference was not detected at competitor densities below 50–100 birds ha−1. Similarly, in mussel-feeding oystercatchers interference only has a significant effect on intake rate above this threshold density (Stillman et al. 1996). Two of the previous studies (Goss-Custard 1977; Sutherland & Koene 1982) were based on bird densities below 100 birds ha−1, and therefore in the density range in which interference is insignificant. The results of both this study and behaviour-based models (e.g. Ruxton et al. 1992; Moody & Houston 1995; Stillman et al. 1997) suggest that the intensity of interference is greatest when prey are scarce. Norris & Johnstone (1998) studied oystercatchers in an area of extremely high cockle abundance in which oystercatchers encountered cockles at a very high rate. Under such conditions, interference would be expected to be relatively week or absent altogether. Ens et al. (1996) studied oystercatchers feeding over a wide range of competitor densities and in areas with a wide range of cockle densities. However, their study was not conducted during late winter when interference in mussel-feeding birds intensifies and this, as Ens et al. (1996) suggest, may be the reason why they failed to detect interference. Therefore, each of the previous studies did not detect interference because some of the essential conditions for interference to occur were missing. In the present study, interference occurred because all of these conditions were present.

The strength of interference is often measured as the slope of the relationship between log intake (or feeding) rate and log competitor density. In mussel-feeding oystercatchers foraging at high competitor densities, this slope usually lies within the range −0·2 to −0·4 (Stillman et al. 1996). In the present study, both feeding rate and competitor density were log transformed in all analyses and so the strength of interference can be compared to that in mussel-feeders. The strength of interference varied both with the food supply and competitor density but ranged from −0·04 (2500 cockles m−2 in November and over 50 birds ha−1) to −0·25 (250 cockles m−2 in November and over 50 birds ha−1) (calculated from Table 3d). The range of interference strength found in cockle-feeders therefore overlaps that found in mussel-feeders but only when the cockle food supply is rare: generally, it is lower than that found in mussel feeders. A possible explanation for this is suggested by behaviour-based interference models (Ruxton et al. 1992; Moody & Houston 1995) which predict that interference will be weaker when handling time is short. The opportunity to steal prey is lower when handling time is short because prey are more often consumed before potential aggressors have a chance to launch an attack. Oystercatchers consuming cockles have a shorter handling time than mussel-feeders and so, even when it is profitable to attempt to steal prey, the reduced opportunity for doing so may reduce the overall strength of interference.

Cockle handling times were negatively related to oystercatcher competitor density. As kleptoparasitism may be less frequent when handling time is short (Ruxton et al. 1992; Moody & Houston 1995; Stillman et al. 1997), this relationship may be evidence that oystercatchers reduce their handling times at high competitor densities in an attempt to reduce their chances of being attacked by a competitor. However, other behavioural responses at high competitor densities, which could also reduce the chances of kleptoparasitism, may increase handling time. In mussel-feeding oystercatchers these include spending an increased proportion of the time scanning for potential aggressors while handling prey (Cayford 1988) and carrying prey away from potential aggressors. Additionally, the negative relationship between handling time and competitor density may have occurred because birds simply aggregated in areas in which cockles could be handled particularly quickly (for example, due to a favourable sediment type). The exact cause of the relationship is therefore unclear. Further studies are required to clarify the relationships between handling time, competitor density and the frequency of kleptoparasitism.

The overall shape of the interference function in cockle-feeding oystercatchers in this study is rather similar to that previously observed in mussel-feeders (Stillman et al. 1996). Interference had a negligible effect on feeding rates at low bird densities, and only decreased feeding rate to any extent at densities above 50–100 birds ha−1. This similarity suggests that there are some common components in interference in the two systems. Stillman et al. (1997) showed that the intensity of interference predicted in mussel feeders was most sensitive to the distance over which interactions between birds occurred. Other parameters, such as handling time and prey encounter rate, which may differ widely for different types of prey, had relatively little impact on the model's predictions. If the distance over which oystercatchers interact is similar in cockles and mussels, a similar form of interference function may be expected. Studies of the interaction distances of oystercatchers, or indeed of any species, feeding on a wide range of prey species may therefore be useful in understanding and predicting the form of interference relationships.

A relationship between prey abundance and the strength of interference has been predicted by behaviour-based models but previous empirical evidence for its existence in birds is rare. However, two previous experimental studies on snow buntings feeding on seed (Dolman 1995) and blackbirds feeding on fat (Cresswell 1998) also found increased interference at low to mid prey densities. Although these studies showed the same pattern as the present study, kleptoparasitism was absent in both snow buntings and blackbirds, and so the mechanisms of interference were different to that observed between cockle-feeding oystercatchers. Interference occurred between snow buntings because dominant birds displaced subdominants from particularly rich microsites and occurred between blackbirds because of the cost of monitoring competitors in order to avoid aggressive interactions. The present study also differs in the manor in which prey abundance varied. The cockle population in the Baie de Somme naturally varied widely both between and within years, whereas the densities of seed and fat available to snow buntings and blackbirds were varied artificially. The present study therefore shows how natural, rather than artificial, variation in food supply can influence the strength of interference between birds.

A relationship between food abundance and interference has not been found in mussel-feeding oystercatchers despite the large number of interference studies in this system (Ens & Cayford 1996). However, mussel densities in the sheltered estuaries in which most studies have been conducted are relatively stable (see, for example, McGrorty et al. (1990) for the population dynamics of mussels in the Exe estuary). It may be that mussel populations do not vary enough to enable a relationship between food supply and interference to be detected. In contrast, the cockle population in the Baie de Somme is highly variable and this may have increased the chance of detecting any relationship. Alternatively, the different combinations of handling time and prey encounter rate for oystercatchers feeding on the two species may be important. Prey stealing is predicted to be profitable, and hence interference to occur, unless handling time is short and prey encounter rate high (Stillman et al. 1997). Mussel-feeders have long handling times and relatively low prey encounter rates and so prey stealing in this system is predicted to be highly profitable (Stillman et al. 1997). Very large changes in these parameters would be needed to make prey stealing unprofitable and so eliminate this cause of interference. In contrast, cockle-feeding birds have shorter handling times and potentially higher encounter rates and so fall within the region in which small changes in either parameter may effect the profitability of prey stealing (Stillman et al. 1997). Changes in the cockle population and associated changes in handling time and encounter rate may therefore be more likely to influence interference in cockle-feeders than in mussel-feeders. More studies of interference in systems with highly variable prey populations are required in order to further understand the relationship between prey abundance and interference.

Acknowledgements

We are very grateful to Sébastien Bacquet and Arnaud Lengignon for help in the field and to Richard Caldow, Ken Norris and Graeme Ruxton for many useful discussions and providing valuable comments on the manuscript. R.A.S. was funded by the Commission of the European Communities and the Natural Environment Research Council.

Ancillary