223

EGG PREDATION IN THE HERRING GULL Larus argentatus: WHY DOES IT VARY SO MUCH BETWEEN NESTS?

ALEWIJN BROUWER & ARIE L. SPAANS

Abstract We studied egg loss in Herring Gulls Larus argentatus on the Dutch Frisian Island of Terschelling. Conspecific egg predation was the main cause ofegg loss. Late breeding, a small inter-nest distance and a low vegetation cover enhanced egg predation. Predation was more frequent in nests with small eggs than in nests with large eggs. The effect of egg vol­ ume on predation was experimentally investigated by exchanging clutches between pairs. Increasing differences in egg size between original and adopted clutches increased progressively the predation rate. Egg predation was strongly correlated to the original clutch volume (control and experi­ mental pairs combined). No correlation was observed between egg preda­ tion and the volume of the adopted clutches. This excludes a possible pred­ ator-linked selection mechanism. It is suggested that lower quality birds not only lay smaller eggs but also exhibit less efficient parental care during the incubation period, resulting in an increased predation risk. DLO Institute for Forestry and Nature Research (IBN-DLO), Department of Animal Ecology, P.O. Box 23, NL-6700 AA Wageningen, The Nether­ lands.

INTRODUCTION We also used the adopted clutches to deter­ mine whether the negative relationship we found In many large Larus gulls, predation of eggs and between predation rate and clutch volume was chicks by conspecifics is a common cause of egg due to predator selection or to the fact that clutch and chick loss (e.g. Brown 1967, Parsons 1971, volume was linked to the care behaviour of the Burger 1984). Cannibalistic behaviour is particu­ parents. In the first case we would expect a corre­ larly severe in areas with a high nest density, thus lation between predation rate and the volume of favouring less densely populated colonies with the clutches as measured after the exchange (new larger territories (Pierotti 1982, Burger 1984). clutch volume). In the second case we would ex­ Since egg and chick predation is often prevalent, pect predation rate to be correlated to the volume it can have important consequences for the total of the clutches as measured before the manipula­ reproductive output of gulls. tion (original clutch volume). If both predation We studied egg predation in Herring Gulls L. selection and care behaviour play a role, we argentatus in a large mixed colony ofHerring and would expect predation rate to be correlated to Lesser Black-backed Gulls L.fuscus in The Neth­ both the new and the original clutch volume. erlands. The study was originally initiated to es­ tablish a causal relationship between laying date and chick survival (Brouwer et al. 1995). For that STUDY AREA AND METHODS study we exchanged complete clutches of three eggs between pairs which had started egg-laying The study was conducted in the Boschplaat nature at different dates. We used the data of egg preda­ reserve (4400 ha) on the eastern part of the Dutch tion on experimental and control clutches to in­ Frisian Island of Terschelling. The reserve con­ vestigate the effects of laying date, vegetation sists primarily of a series of low undulating dune cover, nest density and egg volume on predation complexes surrounded by saltmarshes, which are risk. intersected by large tidal creeks. The dunes are

Received 20 January 1994, accepted 13 June 1994. ARDEA 82: 223.-231 224 ARDEA 82(2), 1994 predominantly covered with marram grass Am­ six days. Exchange took place shortly after the mophila arenaria and sea buckthorn Hippophae last pair of each set of pairs had completed egg­ rhamnoides, with local patches oflyme grass Ley­ laying. In 75 control nests, eggs were removed for mus arenarius, woody nightshade Solanum dul­ approximately five minutes, which was as long as camare and elder Sambucus nigra. The dunes are it took to exchange clutches from experimental interspersed with small grassy valleys dominated pairs. In half of the control nests, eggs were tem­ by sea couch Elymus pycnanthus. Herring Gulls porarily removed just after clutch completion, breed mainly in the dunes, while Lesser Black­ whilst this was done six days later in the other backed Gulls nest predominantly in the valleys. half. Nests were never left empty, but were tem­ The number of Herring Gulls on the island in­ porarily filled with three hard-boiled Herring Gull creased from 6000-8000 pairs in the late 1960s to eggs. Experimental and control nests were chosen 21,000-21,500 pairs in the early 1980s. From 1985 randomly. to 1990 numbers decreased to 15,800 pairs. Ap­ Predation was scored from the moment of ma­ proximately 7 ha of typical nesting habitat used nipulation (exchange or temporary removal of by Herring Gulls was selected as a study area (see clutches). This implies that during the experi­ Spaans et al. 1987, for further descriptive and top­ ment, half of the experimental and control groups ographical details of the study area). was exposed to predation for 26 days (incubation From late April to late June 1990, the study period after the laying period), and half of the area was searched for new clutches in such a way groups for 20 days. So the 167 clutches can be di­ that the entire area was covered at least once eve­ vided into four categories: control clutches ex­ ry two days. During the laying and hatching stag­ posed for 20 (CA) and 26 days (CB), respectively, es, nests were checked daily. Each nest was and experimental clutches exposed for 20 (EA) marked with a small numbered stick as soon as and 26 days (EB), respectively. the first egg was found in the nest. Eggs were in­ For a statistical analysis of the data we used dividually marked a, b, and c, according to their linear and logistic regression with a stepwise laying sequence. Maximum length and width of backward procedure (McCullagh & NeIder 1983). the eggs were measured to the nearest 0.1 mm, The logistic regression is designed to describe using vernier callipers. Egg volume (V in cm3) proportions as a function of one or more indepen­ was calculated using the formula: V = 0.5035· dent variables. The regression gives a maximum length·width2 (Spaans & Spaans 1975). likelihood estimate of the exponent a in the form: Vegetation cover around each nest was scored ea/(l+ea). Significance of variables was tested us­ on a scale from 1 (no, or hardly any cover around ing the F distributed change in deviance and the nest) to 5 (tall vegetation completely covering the change in degrees of freedom when the variable nest). Vegetation cover depended mainly on the was dropped from the full model. In visualising plant species present around the nest; vegetation the relationship between predation rate and each growth during the season had little effect on the significant variable, the effect ofother variables is cover score. At the end of the egg-laying season, not taken into account. the inter-nest distance of each nest was calculated by averaging the distance from the three closest neighbours (Burger 1984). In doing this, we made RESULTS no distinction between Herring and Lesser Black­ backed Gulls. Natural trends For the purpose ofa study on the effect of lay­ We marked 348 clutches of which 167 had ing date on chick survival, 92 clutches of three three eggs at the moment the clutch was ex­ eggs were exchanged between pairs. The differ­ changed or temporarily removed. From the re­ ence in laying date between each set of pairs was maining 181 clutches one or more eggs were Brouwer & Spaans: EGG PREDATION IN HERRING GULLS 225

0.5 • 0.5

0.4 0.4 • c: a .2 0.3 • a Cil "0 § 0.3 ~ 0.2 ~ ~ ----(J--o--- ~ a. "0 • I •• ~ 0.1 0.2 a • • 0.0'-----7-----=-----:!:----~--____!:_---' 234 5 0.1 cover

0.5 0.0 220 230 240 250 260 270 280 c1uch volume (cm3) 0.4 • • c: • .2 0.3 Fig. 2. Relationship between predation and clutch Cil • "0 volume in Herring Gulls. Black dots (each dot repre­ ~ 0.2 a. •• senting 23-24 nests) include original eggs of experi­ 0.1 mental clutches and the eggs of the control group (N = • 165). Open dots (each dot representing 12-13 nests) in­ 0.0 2 • clude eggs of nests of the experimental group after the 4 6 8 10 12 inter-nest distance exchange (adopted eggs: N = 90).

0.5 0.4 • From the 167 complete clutches 105 eggs (25 c: • • times three eggs, 3 times two eggs, 24 times one .Q 0.3 • Cil egg) disappeared after the exchange or temporary "0 ~ 0.2 removal of the clutch. The chance of losing an a. egg increased progressively from 0.14 for the first 0.1 egg within a clutch to 0.89 of also losing the third 0.0 egg. Within the group ofpartially robbed clutches 5 10 15 20 25 30 laying date predation was random with respect to the laying sequence of the eggs CX2= 0.85, df= 2, NS). Fig. 1. Herring Gull egg predation in relation to cov­ Using logistic regression, a model was con­ er (upper graph), inter-nest distance (central graph) and structed for the 167 clutches that had three eggs at laying date (lower graph). Predation is expressed as the the moment of manipulation. The model de­ proportion of eggs lost per clutch of three eggs. Inter­ scribed the relationship between the proportion of nest distance is the average distance to the three nem;est eggs lost per nest, and laying date, vegetation neighbours. Each dot represents 14 nests. Both control cover, inter-nest distance and original clutch vol­ and experimental pairs are included (for statistics see ume (volume of clutches before manipulation of Table 1). the eggs). Each of these four variables significant­ ly decreased the deviance (Table 1). The probabil­ preyed upon by conspecifics before the planned ity of egg predation decreased with an increase in moment of manipulation. The mean volume of cover, inter-nest distance and original clutch vol­ the a-eggs of these clutches differed significantly ume (Figs 1 & 2). from the mean volume of the a-eggs of clutches Because cover and inter-nest distance in­ that retained their eggs (85.5 and 88.7 cm3, re­ creased with laying date (linear regression, cover: 2 spectively, t test, t166 =2.92, P < 0.01). 0.0485·laying date + 2.18, r 165 = 0.08, P < 0.001; 226 ARDEA 82(2),1994

Table 1. Logistic regression analysis of the proportion of Herring Gull eggs lost per clutch of three eggs for con­ trol and experimental pairs. The null model represents the constant only, the full model includes all significant pre­ dictor variables. For the null and full model the deviance and the degrees of freedom are given, for the significant variables the F distributed change in deviance and the change in degrees of freedom are listed after stepwise drop­ ping variables from the full model.

Parameter F (dev.) Degrees of Estimates of freedom 1 coefficients

Null model 411.3 166 Full model 314.9 158 Laying date 9.60** +1 0.1094 Vegetation cover 9.13** +1 -0.6306 Original clutch volume 8.34** +1 -0.0856 Inter-nest distance 4.97* +1 -3.4090 Clutch volume * inter-nest distance 4.22* +1 0.0123 Squared difference of clutch volumes 4.77* +1 0.0005

1* P < 0.05, ** P < 0.01

inter-nest distance: 0.0767·1aying date + 3.83, 0.5 2 r 165 = 0.033, P < 0.05), a decrease in egg preda­ tion with laying date would be expected. Howev­ 0.4 er, egg predation also increased with laying date (Fig. 1, Table 1). g 0.3 Considering all possible interactions, only the 'iii "0 Ql interaction between original clutch volume and a. 0.2 o inter-nest distance was significant (Table 1). The ...... effect of clutch volume on predation rate in­ 0.1 ...... creased with decreasing inter-nest distance (Fig. o ...... 3). All other two-way interactions between the -­ 0.0 220 230 240 250 260 280 parameters mentioned in Table 1 were not signifi­ cluch volume (cm 3) cant. To test non-linear relationships all parame­ ters were squared and proven not to be significant Fig. 3. Relationship between predation and clutch either. volume in Herring Gulls at two different classes of in­ ter-nest distances (open dots: 6.53 m, SD = 2.68, N = The effect ofclutch transfer 82, black dots: 3.69 m, SD = 0.62, N = 83). Each dot As a next step the two experimental categories represents 20-21 nests. (EA, EB) and the two control categories (CA, CB) were added to the full model, as categorical In the experimental pairs, new and original variables. Neither of these variables, nor their clutch volume differed from each other in all cas­ interactions significantly decreased the deviance. es; in the control pairs new and original clutch This implies that within the model there were no volume were of course identical. The squared dif­ significant differences in predation rate between ference between the original and new clutch vol­ the four categories. umes (dvoI2) significantly decreased the deviance Brouwer & Spaans: EGG PREDATION IN HERRING GULLS 227

0.5 predation and new clutch volume (Fig. 2). Con­ trolling for new clutch volume, the interaction 0.4 term new clutch volume X exchanged significant­ ly decreased the deviance (F l161 = 4.79, P < p~irs § 0.3 0.05). Using the experimental only, the coef­ .~ ficient of the new clutch volume did not signifi­ "0 CD 0. 0.2 cantly differ from zero (Fl ,88 = 1.0, NS). Controllirtg for original clutch volume, the 0.1 interaction term original clutch volume X ex­ changed did not significantly decrease the devi­ 0.0 ance (F1 161 = 2.18, NS), indicating that in the re­ control E1 E2 E3 lationship between predation and origirtal clutch volume, the experimental pairs follow the same Fig. 4. Mean predation rate of control (light grey, trend as the control pairs. From this we conclude N = 75) and experimental clutches of Herring Gulls that the original clutch volume affected the preda­ (£1-£3). Experimental clutches with increasing squar­ tion rate, whilst the new clutch volume did not. ed difference between the volume (cm3) of the original clutch and that of the adopted clutch (dvoI2: £1 = This means that predation rate is not directly re­ 56.25, SD = 50.57, N = 30, £2 = 364.6, SD = 161.6, N = lated to egg size. 30, £3 = 1479, SD = 1020, N = 30).

(Table 1). Predation risk increased with an in­ DISCUSSION creasing difference between original and new clutch volume (Fig. 4). As a result the experimen­ Egg predation by conspecifics was a common tal pairs (EA, EB) suffered a higher predation rate phenomenon in Herring Gulls on Terschelling irt than the control pairs (nests: 39.2% and 21.3%, 1990, both during the layirtg and the incubation respectively; eggs: 27.2% and 13.3%, respective­ period. The high predation rate irt 1990 was not ly). The effect of dvol2 was not significant when exceptional. In 1983-1984, almost a quarter of all the difference between original and new clutch eggs produced was lost to predators (Spaans et al. volume was investigated for the 92 experimental 1987). pairs only (F1•83 = 3.10, NS). Once a clutch was fully incubated, egg preda­ tion was random with regard to the laying se­ Clutch volume investigated experimentally quence of the eggs. This contrasts with a study In the full model (Table 1), we then replaced conducted by Verbeek (1988), who found a selec­ the variable original clutch volume with the new tive predation for the c-egg in the Glaucous-wing­ clutch volume for control and experimental pairs ed Gull L. glaucescens in Canada. In Verbeek's combined. In contrast to the origirtal clutch vol­ study, however, the c-egg was lighter coloured ume, the coefficient of the new clutch volume did than the two other eggs, thus focussing the atten­ not significantly differ from zero (F1.159 = 2.97, tion of predators on that particular egg (see also NS). Verbeek 1990). On Terschelling there were no Finally, we constructed a new model irtclud­ consistent colour differences between the eggs ing only the variables new clutch volume and withirt a clutch. 'exchanged' (categorical) and used the interaction The probability of losing another egg, once a term new clutch volume X exchanged to establish pair fully irtcubating a c/3 had lost an egg, in­ a difference in slope between the original clutches creased progressively. The same result was found (control and experimental pairs combirted) and experimentally by Baerends et al. (1970), who the adopted clutches irt the relationship between showed that a progressive egg loss is linked with 228 ARDEA 82(2), 1994 an increase in restlessness of the incubating bird unknown whether the reluctance ofearly breeders and in a tendency to leave the nest more often. to nest in tall vegetation has evolved under The rate of predation increased in later laying present predation pressure or that is has an ulti­ birds, whilst the probability of losing eggs de­ mate basis. creased with an increase in vegetation cover, in­ If conspecific egg predation is so negatively ter-nest distance and clutch volume. The relation­ correlated to inter-nest distance, one may ask why ship to clutch volume was absent after clutches gulls do not nest more solitarily. Eggs placed by had been exchanged between pairs. Size differ­ Burger (1984) in unattended experimental nests ences between original and adopted eggs in­ outside the colony, were, however, faster found creased the predation rate progressively. Differ­ and eaten by predators than eggs placed in the ences in contact between eggs and brood patch colony. This suggests that neighbouring gulls, before and after the exchange may have induced though potential predators themselves, serve at an increased restlessness in the experimental least partially as a deterrent to potential robbers birds, thus enhancing the increase in predation roaming the area. However, a tendency for soli­ rate. Besides size differences, other tactile stimuli tary nesting as a result of severe egg loss has been and visual features of the eggs could also have described for the Yellow-footed Gull L. livens by contributed to the effect of the manipulation (see Spear & Anderson (1989). Baerends & Hogan-Warburg 1982). An increase in egg predation with an increas­ Since eggs are less easily seen as the vegeta­ ing nest density was also noticed by Fordham tion becomes taller (Brown 1967, Burger 1984) the (1964) for the Southern Black-backed Gull L. negative relationship we found between vegeta­ dominicanus. Parsons (1976) and Burger (1984), tion cover and predation rate was as expected. however, found the lowest predation rate in Her­ Later laying birds preferred progressively a tall ring Gull territories of intermediate sizes. Burger vegetation such as sea buckthorn. If a tall vegeta­ (1984) argued that at a high nest density the gulls tion is advantageous from the egg predation point got too involved in neighbour-neighbour interac­ of view, the question arises as to why early gulls tions, resulting in a waning of egg guarding. do not choose nesting sites in this habitat more of­ However, if the nest density was too low, gulls ten. We suggest that nesting in a tall vegetation had frequent encounters with non-neighbour in­ increases the adult's risk of being killed during truders that were attempting to establish a territo­ laying and incubation by a mammalian predator. ry. Gulls with territories of intermediate sizes had This is strengthened by the observation of Drent the lowest aggressive interaction rate and there­ (1970) on nearby Schiermonnikoog that Herring fore the highest reproductive output. The relation­ Gulls nesting in tall and dense vegetation aban­ ship between nest density and interaction rate don their eggs and the colony during darkness for with other gulls is, however, habitat dependent many more days after clutch completion than (Pierotti 1987). It is therefore difficult to make birds nesting in open habitat. As the season pro­ generalisations. gresses an increasing number of gulls remain in We found experimental evidence that within the colony during darkness, thus probably facili­ clutches of three eggs predation of eggs is not tating later starting gulls to nest more in tall vege­ causally linked to clutch volume. Apparently, tation. Mammalian predators are lacking on Ter­ there is no predator selection with respect to schelling and Schiermonnikoog except for small clutch volume. The relationship between original numbers of stoats Mustela erminea (Terschelling clutch volume and predation rate must have been 1930s-1980s, Schiermonnikoog 1980s, see Broek­ caused therefore by a factor correlated to clutch huizen et al. 1992) and feral cats Felis catus (last volume. Egg volume is directly affected by food few decades, State Forestry Service Terschelling supply (Hiom et al. 1991, Van Klinken 1992) and and EJ.J. Niewold pers. comm.). It is therefore the physiological state of the female (Pierotti & Brouwer & Spaans: EGG PREDATION IN HERRING GULLS 229

Bellrose 1986, Meathrel et at. 1987). Mills (1979) REFERENCES showed that the most efficient foragers produce the largest eggs, suggesting that parental quality Baerends, G.P., R.H. Drent, P. Glas & H. Groenewold is related to egg size. A positive relationship 1970. An ethological analysis of incubation beha­ viour in the Herring Gull. Behaviour, Supplement between egg size and parental quality was clearly 17:135-235. established by Bolton (1991), who performed a Baerends, G.P. & A.J. Hogan-Warburg 1982. The ex­ clutch transfer experiment in Lesser Black­ ternal morphology of the egg and its variability. backed Gulls by cross-fostering clutches of large Behaviour 82: 1-32. eggs with clutches of small eggs. We suggest Bolton, M. 1991. Determinants of chick survival in the therefore that lower quality birds not only lay Lesser Black-backed Gull: relative contributions of egg size and parental quality. J. Anim. EcoL smaller eggs as a result of less efficient foraging 60:949-960. practices but also show less efficient egg care be­ Broekhuizen, S., B. Hoekstra, V. van Laar, C. Smeenk haviour (e.g. lower territory attendance) during & J.B.M. Thissen 1992. Atlas van de Nederlandse the incubation period, resulting in an increased zoogdieren. Stichting Uitgeverij Koninklijke Ne­ derlandse Natuurhistorische Vereniging, Utrecht. predation risk. We have no data on territory atten­ Brouwer, A., A.L. Spaans & A.A.N. de Wit 1995. Sur­ dance or other parental care behaviour in relation vival of Herring Gull Larus argentatus chicks: an to clutch volume. However, Herring Gulls that experimental analysis of the need for early breed­ lose eggs, more frequently leave the colony for ing. Ibis 137, in press. feeding (probably as a result of a lower hunting Brown, RG.B. 1967. Breeding success and population growth in a colony of Herring and Lesser Black­ efficiency) than successful pairs. The former also backed Gulls Larus argentatus and Larus fuscus. leave their territory unguarded for longer periods Ibis 109:502-515. of time than the latter (M. Bukaciilska & D. Bu­ Burger, J. 1984. Pattern, mechanism, and adaptive sig­ kaciilsky, pers. comm.). These data indicate that nificance of territoriality in Herring Gulls (Larus parental care behaviour and breeding success on argentatus). OrnithoL Monogr. 34:1-92. Burger, J. 1987. Selection for equatability in some as­ Terschelling are strongly related to each other. pects of reproductive investment in Herring Gulls Burger (1987) and Morris (1987) came to the same Larus argentatus. Ornis scand. 18:17-23. conclusion for Herring Gulls nesting in North Drent, R.H. 1970. Functional aspects of incubation in America. the Herring Gull. Behaviour, Supplement 17: 1­ 132. Fordham, RA. 1964. Breeding biology of the Southern Black-backed Gull II: Incubation and the chick ACKNOWLEDGEMENTS stage. Notornis 11:110-126. Hiom, L., M. Bolton, P. Monaghan & W. Worrall 1991. We would like to thank Harry Hom, State Forestry Ser­ Experimental evidence for food limitation of egg vice on Terschelling, for permission to conduct this production in gulls. Ornis scand. 22:94-97. study in the Boschplaat nature reserve, and the bird McCullagh, P. & J.A. NeIder 1983. Generalised linear wardens Leo Bot, Jan Bunnink, Oene de Jong, Jan models. Chapman & Hall, London. Peereboom, Jan-Piet Rijf, and Jan Vis for logistic sup­ Meathrel, C.E., J.P. Ryder & B.M. Termaat 1987. Size and composition of Herring Gull eggs: relation­ port in the field. Wilfred Hayer was an indefatigable ship to position in the laying sequence and the field assistant. We thank Martin Brinkhof and Simon body condition of females. Colonial Waterbirds Verhulst for introducing us to the logistic regression 10:55-63. analysis. Dr. Hileo van der Voet provided additional Mills, J.A. 1979. Factors affecting the egg-size of Red­ statistical support. Miriam Hall corrected the English billed Gulls Larus novaehollandiae scopulinus. text. We are also grateful to Dr. Mark Bolton, Martin Ibis 121:53-67. Brinkhof, Monika Bukacinska, Dariusz Bukacinski, Si­ Morris, RD. 1987. Time-partitioning of clutch and mon Verhulst, and our colleagues at the Institute for brood care activities in Herring Gulls: a measure helpful comments on earlier drafts of the paper. ofparental quality? Studies Avian BioI. 10:68-74. Parsons, J. 1971. Cannibalism in Herring Gulls. Brit. Birds 64:528-537. 230 ARDEA 82(2), 1994

Parsons, J. 1976. Nesting density and breeding success SAMENVATTING in the Herring Gull Larus argentatus. Ibis 118: 537-546. Zilvermeeuwen Larus argentatus roven geregeld eie­ Pierotti, R. 1982. Habitat selection and its effect on re­ ren uit nesten van soortgenoten. De kans op eierroof is productive output in the Herring Gull in New­ foundland. Ecology 63:854-868. niet voor ieder nest gelijk. In 1990 hebben wij op Ter­ Pierotti, R. 1987. Behavioral consequences of habitat schelling onderzocht welke factoren van invloed zijn selection in the Herring Gull. Studies Avian BioI. op deze eipredatie. Uit 167 legsels met drle eieren 10:119-128. werden in totaal 105 eieren geroofd. AIle eieren binnen Pierotti, R. & C.E. Bellrose 1986. Proximate and ulti­ een legsel hadden een even grote kans om te worden mate causation of egg size and the "third-chick gepredeerd. Als er eenmaal een ei uit een nest was ge­ disadvantage" in the Western Gull. Auk 103:401­ roofd, nam de kans op verdere roof toe. 407. De kans op eipredatie nam af naarmate de vegetatie Spaans, A.L., AAN. de Wit & M.A van Vlaardingen rond het nest hoger was en de nesten verder uit elkaar 1987. Effects of increased population size in Her­ lagen. Naarmate het seizoen vorderde, gingen de meeu­ ring Gulls on breeding success and other parame­ ters. Studies Avian BioI. 10:57-65. wen steeds verder uit elkaar en in een hogere vegetatie Spaans, M.J. & AL. Spaans 1975. Enkele gegevens nestelen. Desondanks nam de eipredatie toe naarmate over de broedbiologie van de Zilvermeeuw Larus de meeuwen later begonnen te leggen. argentatus op Terschelling. Limosa 48:1-39. Nesten met kleine eieren hadden een grotere kans Spear, L.B. & D.W. Anderson 1989. Nest-site selection om gepredeerd te worden dan nesten met grote eieren. by Yellow-footed Gulls. Condor 91:91-99. Door legsels te verwisselen voor legsels van soortgeno­ Van Klinken, A 1992. The impact of additional food ten, kon worden aangetoond dat er geen direct verband provisioning on chick growth and breeding output bestaat tussen eigrootte en de kans op predatie. Veron­ in the Herring Gull Larus argentatus: a pilot ex­ dersteld wordt dat eigrootte en broedzorg beide in periment. Ardea 80:151-155. sterke mate worden bepaald door de kwaliteit van de Verbeek, N.AM. 1988. Differential predation of eggs in clutches of Glaucous-winged Gulls Larus glau­ oudervogel. cescens. Ibis 130:512-518. Verbeek, N.AM. 1990. Differential predation on eggs in clutches of Northwestern Crows: the impor­ tance of egg color. Condor 92:695-701.