Factors Influencing Rate and Success of Intraspecific Kleptoparasitism Among Kelp Gulls (Larus Dominicanus)

The Auk 112(4):847-859, 1995

FACTORS INFLUENCING RATE AND SUCCESS OF INTRASPECIFIC KLEPTOPARASITISM AMONG KELP GULLS (LARUS DOMINICANUS)

WILLIAM K. STEELE AND PHILIP A. R. HOCKEY PercyFitzPatrick Institute of AfricanOrnithology, University of CapeTown, Rondebosch,South Africa 7700

AI•STRACT.--Intraspecifickleptoparasitism among Kelp Gulls(Larus dominicanus) was studied at four sitesin the southwesternCape Province of South Africa. Sitesincluded: (1) undisturbed foraging habitats;(2) a sandybeach and a rocky shore;and (3) areaswhere supplementary food was available--a fishing harbor and a refuse dump. Simple food-choiceexperiments were usedto test hypothesesgenerated from field observations.Among-site variation in the rate and successof kleptoparasitismwas related to prey attributes,of which prey size and handling time were the most important. In food-choiceexperiments, gulls selectedsmall prey with short handling times. Prey with long handling times were the most likely to be stolenand the rate of keptoparasitismwas higher when prey were dispersedthan when they were clumped. There were marked age-related differencesin the rate, although not the success,of kleptoparasitismamong Kelp Gulls. Juvenile(first-year) gulls attemptedklepto- parasitismsignificantly more often than expectedand adults significantly less often. Subadults kleptoparasitizedin proportion to their abtmdancein the population.If an age-relateddom- inancehierarchy exists, it mediateskleptoparasitic behavior in Kelp Gull assemblagesthrough older birdsavoiding kleptoparasitic attacks rather than initiating them. Simple mathematical models, based on data collected during field observations,were used to investigate the conditionsexplaining the rateof intraspecifickleptoparasitism within Kelp Gull populations. Either few individuals can kleptoparasitizerelatively frequently, or many individuals can kleptoparasitizeinfrequently. Apparently, both mechanismsoperate within Kelp Gull populations becauseindividuals attempt kleptoparasitismrelatively frequently when they are juvenilesand inefficienthunters, but infrequently oncethey are adult and efficienthunters. The viability of facultativeintraspecific kleptoparasitism as a foraging techniquerelies on stolenprey being larger on averagethan the prey capturedby hunting. Received18 July1994, accepted13 February1995.

KLEPTOPARASITISM,or the stealing of already populationscan be controlled in two ways: ei- procuredprey from one individual by another ther few individuals kleptoparasitizerelatively (Brockmann and Barnard 1979), has been ex- frequently, or many individuals kleptoparasi- tensively studied where it occursinterspecifi- tize relatively infrequently. Alternatively, in- cally among birds (see reviews by Brockmann dividuals can changetheir role aseither a "pro- and Barnard1979, Barnard and Thompson1985, ducer" or a "scrounger,"depending on the na- Furness 1987). Intraspecific kleptoparasitism ture of the food resourceor the compositionof among birds has received lessattention in the the foraging group (Giraldeau and LeFebvre scientificliterature, but field evidenceof itsprof- 1986). Although differential frequencies of itability or otherwise has been discussedby kleptoparasitismamong age classeswithin gull Kushlan (1978, 1979), Dunbrack (1979), Hansen populations have been shown (Burger and (1986), and Ens et al. (1990). Gochfeld 1979, Carrol and Cramer 1985, Hesp Intraspecifickleptoparasitism differs from the and Barnard 1989, Hockey et al. 1989, Hockey interspecificbehavior becausethe rate of intra- and Steele 1990), the constraints that determine specifickleptoparasitism within a populationis this division are poorly understood. constrainedby factors that do not operate in- The Kelp Gull (Larus dominicanus)is wide- terspecifically(Barnard and Sibly 1981,Vickery spreadthroughout the higher latitudesof the et al. 1991). Some individuals must hunt, and Southern Hemisphere (Harrison 1983), and is all individualscannot resort to kleptoparasitism common along much of the southern African as a foraging technique,except at a very low coastline (Crawford et al. 1982, Steele and Hock- rate. Thus, intraspecifickleptoparasitism within ey 1990). These gulls are separable into age 847 848 STEELEAND HOCKEY [Auk, Vol. 112 classesin the field on the basisof plumage and 18ø07'E)and Bloubergstrand(33ø47'S, 18ø27'E), where soft-part coloration (Kinsky 1963, Harrison all interactionsbetween gulls were clearly visible. At 1983), allowing age-related differencesin for- Kalk Bay harbor,a food-choiceexperiment was used aging and kleptoparasiticbehavior to be stud- to test the effectsof prey size on food selection and kleptoparasiticbehavior. Three piecesof fish offal of ied. Field observationsof flocksof foragingKelp different size classes(small, ca. 25 x 25 x 50 mm; Gulls, supported by food-choiceexperiments, medium, ca. 25 x 25 x 100 mm; large, ca. 25 x 25 x were used to determine the factors that influ- 150mm) were placedequidistant from the main group ence the rate and successof intraspecificklep- of gulls. The order in which the food items were toparasitism. placedwas varied eachtime (n = 50 replicates). Becauseit wasimpossible to quantify the full It haspreviously been determinedthat Kelp Gulls foragingrepertoire of all gulls, a mathematical rapidly crushand swallowsmall sand mussels (Donax model was used to estimatethe proportion of serra),whereas large Donaxrequire a complexhan- foraging time an individual should devote to dling technique,in which the gulls fly up and drop kleptoparasitismif all birds were facultative the musselto break the shell open (Steele 1992). In order to test the effectof prey-handling time on food producers/scroungers.A second, empirically- selection, Kelp Gulls at Bloubergstrand were pre- based model was used to investigate the pro- sentedsimultaneously with two large Donaxof equal portion of birds in flocksof different sizesthat size, but one having had its adductor muscle cut so could be obligate kleptoparasites (and still that the valves gaped open slightly and gulls could maintain an intake rate at least equal to pro- gain direct accessto the flesh (n = 51 replicates). ducers). That is, what would be the stable ratio In order to test effectsof prey "quality" (defined of producersto scroungersif all individualsused as energy content per unit mass)on kleptoparasitic only one or the other foraging technique. behavior,independent of prey size,gulls at Kalk Bay harbor were presentedsimultaneously with 10 por- METHODS tions each of fish offal and bread of the same size (ca. 25 x 25 x 25 mm). Breadwas dampenedto make it Field observations.--Threeage classesof Kelp Gull easier for gulls to swallow. These food items were were recognizedon the basisof plumage and bare- placed alternatively in a grid pattern, over 1.5 x 1.0 part coloration:juvenile (first-year birds), subadult m, to ensurethat gulls could selectthe item of their (second-and third-year birds), and adult. Four sites choiceimmediately on landing (n = 34 replicates). in the WesternCape Provinceof SouthAfrica, where Finally, to test the effectsof prey dispersioninde- flocksof Kelp Gulls regularly foraged,were selected pendent of prey size and quality, Kelp Gulls were to include a range of feeding habitats. Olifantsbos videotapedat Hout Bayharbor while feedingon pil- Bay (OB; 34ø16'S,18ø23'E) is a stretch of undisturbed chards(Sardinops ocellata) presented as one pile of 20 rocky shore in the Cape of Good Hope Nature Re- fishes(n = 10 replicates),five piles of 4 (n = 10),and serve, and 16 Mile Beach(SMB; 33ø18'S,18ø17'E) is a 20 individually spacedfishes (n = 10).All fisheswere sandy beach within the West Coast National Park of approximatelythe samesize, and all presentations where humans have had little impact. By contrast,at were evenly spacedover a 10 x 10 m area.The order both Strandfontein refusedump (SD: 34ø05'S,18ø30'E) in which different presentationswere used varied and Hout Bay fishing harbor (HB; 34ø04'S,18ø21'E), randomly, and no more than three replicateswere supplementaryfood is availableto the gullsas a result presentedon any given day. of human activities. During controlled feeding, the age ratio of Kelp We conducted 220 h of observations on flocks of Gulls presentwas recordedat 30-min intervals,and foraging gulls at the four sites from 1986 through the order of food selection, as well as the rate and 1989.The numbers of gulls present during observa- successof kleptoparasiticattacks, were recorded by tion periods were counted hourly by age class.For age class.A secondand smaller gull species,Har- each observedattempt at kleptoparasitism,the suc- tlaub'sGull (L. hartlaubii),was also present at Kalk Bay cess,age classof host, number and age class(es)of harbor during controlledfeeding and participatedin pursuer(s),and age classesof successfulkleptopar- the food-choiceexperiments. The mean number of asiteswere noted. Prey-handling times and klepto- gulls presentat Kalk Bay harbor during controlled parasiticpursuits were timed to the nearest second. feeding was 12 Hartlaub'sand 30 Kelp gulls (n = 7). An index of the rate of kleptoparasitismamong gulls Although Hartlaub'sGulls alsowere presentat Hout at different sites was calculated as the number of ob- Bayharbor, controlled feeding took placeon the har- servedkleptoparasitic incidents per hour for each 100 bor breakwater,which wasthe main Kelp Gull loafing gulls present. area,and Hartlaub'sGulls were totally excludedfrom Food-choiceexperiments.--To test trends observedin the food by the larger gulls. the field, controlled feeding of gulls was carried out Models.--During observations at Strandfontein on open sandybeaches at Kalk Bay harbor (34ø09'S, dump, individual Kelp Gullsscavenging for food were October1995] KleptoparasitismAmong Kelp Gulls 849 selectedat random and watchedfor 5 min, until they culatedreward for both strategieswas comparedfor went to the loafing area, or were lost to sight. The various flock sizes and compositions.For each flock sizeof food itemswas estimatedrelative to bill length size, the composition(ratio of scavengersto klepto- (ca. 54 mm; Maclean 1993) during both scavenging parasites)at which point no individual could better and kleptoparasiticattempts, and a number of other its food intake by changingits foragingstrategy was foraging variables were determined (Appendix 1). determined. These variableswere then used to model aspectsof the foragingbehavior of Kelp Gulls at the dump. Kelp RESULTS Gulls at Strandfontein dump theoretically have a choice among three foraging "strategies":(1) the FACTORS INFLUENCING RATE AND "scavenger"does not resort to kleptoparasitism,but SUCCESS OF KLEPTOPARASITISM scavenges/huntsfor food;(2) the "kleptoparasite"ob- tainsfood through kleptoparasitismalone; and (3) the There were many more Kelp Gulls at sites "scavenger-kleptoparasite"employs a combinationof the two foragingtechniques. Using the Strandfontein where supplementaryfood was available than data,we developedtwo modelsto find the stablerate at undisturbed Sites (Table 1). The number of of kleptoparasitismdepending on whether the be- Kelp Gulls at each study site actively foraging havior was facultativeor obligate. every hour was much lower than the total num- Model I assumesthat all Kelp Gulls in the flockare ber of gulls present. The mean number of for- able to vary the proportionof availableforaging time aging Kelp Gulls at Olifantsbos Bay during they devote to kleptoparasitism(i.e. the flock com- hourly countswas 7.3 + SD of 8.0 (n = 27) and prises scavengersand/or scavenger-kleptoparasites). at Strandfonteindump was 26.9 + 21.9 (n = 33). Individual reward, in terms of the equivalent number The rtumberof gulls foragingat 16 Mile Beach of average-sizedfood items consumedper hour, was and Hout Bay harbor fluctuated so widely calculatedfor a rangeof ratesof kleptoparasitism,and the theoreticalproportion of foraging time that Kelp throughout the day according to tidal condi- Gulls at Strandfontein dump should devote to klep- tions and/or harbor activity that it was mean- toparasitismin order to maximize their rate of food inglessto calculatean averagevalue. The in- intake was determined. The parameters,assumptions, dexesof ratesof kleptoparasitismvaried greatly and calculations used for this model are detailed in betweensites (Table 1), being highestat 16 Mile Appendix 1. Beachand lowest at OlifantsbosBay--the two The casewhere flocksof Kelp Gulls at the Strand- undisturbedstudy sites. Many attemptsat klep- fontein dump comprisea mixture of obligate scav- toparasitism took place among gulls on the engers and obligate kleptoparasiteswas considered in Model II. A model similar to that of Barnard and ground, and theseoften were impossibleto ob- serve when large numbers of gulls were feed- Sibly (1981) was developed to determine the stable ratio of kleptoparasitesto scavengerswithin Kelp Gull ing together. Furthermore, it was only possible flocksof varying size. Individual reward, in terms of to follow a single kleptoparasiticattempt at a the number of average-sizedfood items consumed time and, therefore, the number of recorded per hour, was calculatedfor gulls following one of kleptoparasiticincidents is an underestimate. these two foraging strategies(Appendix 2). The cal- Prey size and handlingtime.--Prey-handling

T^ni•E1. Summary of feeding and kleptoparasiticbehavior of Kelp Gulls at study sites.

Undisturbed Food-supplemented OB SMB HB SD

Mean number at site a 23.4 + 26.3 (17) 24.7 + 22.4 (10) 243 + 150 (17) 546 + 296 (19) Site surface area (m 2) 2,000 10,000 25,000 25,000 Principal prey Dipteran larvae gonax serra, Engrauliscapen- Domestic waste and pupae invertebrates sis,invertebrates Mean prey-handlingtime (s)a <1.0 175.6b + 129.4 (52) 2.1 + 0.4 (8) 3.8 + 11.9 (40) Mean pursuit time (s)a 4.0 (1) 5.0 + 11.4 (30) 2.8 + 1.8 (12) $.3 + 11.4 (138) Kleptoparasitism rate (incidents.[100 gulls]-'. h-') 0.1 6.8 0.7 1.2 Individual kleptoparasitic success (%) 0.0 25.5 15.2 14.0

ß• _+SE (n). b For Donaxonly. 850 STEELEAND HOCKEY [Auk,Vol. 112

TABLE2. Effectof prey size on food choice,and fre- prey item first significantly more often than quencyand success of kleptoparasitismby Kelp and expectedby chance(X 2 = 4.2, df = 1, P < 0.05); Hartlaub's gulls. the largestitem was usually selectedlast (Table Prey size 2). This order of selection reflects the risk of (n = 50 for each size)a losingprey to kleptoparasites:63.4% of all large Small Medium Large food items were stolen, comparedto 29.2 and 6.0% of medium-sized and small-sized items, Order of selection (%) respectively. First 46 42 10 Second 30 46 16 Prey-handlingtechnique.--The handling tech- Third 24 8 56 nique used for Donax at 16 Mile Beach,where Not taken 0 4 18 prey are dropped during flight, made prey Percentagetaken by readily availableto potential kleptoparasites.At Kelp Gull adult 24 42 34 this site, Kelp Gulls also fed on three-spotted Kelp Gull subadult 24 47 29 swimming crabs (Ovalipespunctatus), which re- Kelp Gull juvenile 43 23 34 quired a long handling period (sometimes> 10 Hartlaub's Gull 75 20 5 rain). However, in contrast to the handling Probability of retaining prey by techniqueused for Donax,gulls feeding on Ova- Kelp Gull adult 1.00 0.75 0.22 lipesstood over, and were able to protect their Kelp Gull subadult 1.00 0.50 0.33 prey. As a result, no crabs were lost to klepto- Kelp Gull juvenile 0.87 0.87 0.67 Hartlaub's Gull 0.93 0.50 0.00 parasites. During feeding experiments, Kelp Gulls at ' Small,25 x 25 x 50 ram;medium, 25 x 25 x 100turn; large, 25 x 25 x 150 ram. Bloubergstrandselected partially openedDonax beforesimilarly sizedDonax, which were closed (41 of 51 occasions,X 2 = 18.8, P < 0.01). Prey qualityand abundance.--Itis difficult to time affected both the overall rate (Table 1) and comparerelative prey "quality" between sites successof kleptoparasitismamong gulls. At Oli- in the field. When offered prey of similar size fantsbosBay, where the mean prey-handling and handling time, but of different energetic time was less than 1 s, the rate of kleptopara- content, gulls clearly selected "high-quality" sitism was very low and restrictedto uncom- prey (Table 3). Fish offal (19.0 kJ/g; Hockey mon,large prey items(e.g. mussels, limpets and unpubl. data) was selected before bread (11.1 fishes).At 16 Mile Beachthe principal prey of kJ/g; N.R.I.N.D. 1986) significantly more often Kelp Gulls was Donax,which required a long than expected (X 2 = 16.0, P < 0.01). Fifteen handling time; at this site the rate of klepto- kleptoparasiticattacks were directed at hosts parasitismwas higher than at any other (Table with fish and only one at a host with bread; 1). At Strandfontein dump, successfulklepto- gulls directed attacksdisproportionately more parasiticattempts involved significantlylarger frequently at hostswith high-quality prey (X2 prey (œ = 75 _+ 52 ram, n = 68) than failed = 4.9, P < 0.05). Only 1.2%of fish piecesoffered attempts(œ = 61 _+41 ram, n = 146;t = 2.07, df were left at the end of experimental runs com- = 212, P < 0.05). pared to 53.5% of the piecesof bread. At Kalk Bay harbor, when presented with At Strandfontein dump, food was usually evenly spacedprey of three different size class- abundantand comparativelyfew gulls resorted es simultaneously,gulls selectedthe smallest to kleptoparasitism(Table 1). Although supplementary food also was available at Hout Bay harbor,this was only for limited periods,when TABLE3. Effect of prey quality (in terms of energy boats unloaded catches or cleaned nets. The contentper unit mass)on order of food choice(n = 340 for bread and for offal) by Kelp and Hart- highest observedrate of kleptoparasitismwas laub's gulls. at 16 Mile Beach,where large Donax afforded an opportunity for kleptoparasites.These mus- Fish selswere relatively uncommon,with only one Order of selection Bread offal Unknown or two being handled by gulls at any one time. First item taken a 14.7 82.4 2.9 The successof kleptoparasiticattempts by Kelp Second item taken a 5.9 82.4 11.7 Gulls at 16 Mile Beachwas high in comparison ßPercent of time particularitem taken. to the other sites (Table 1). October1995] KleptoparasitismAmong Kelp Gulls 851

TAnrE4. Effect of prey dispersionon rate of klep- 35O toparasitismamong Kelp Gulls (n = 10 for each column). 300 1 pile of 5 piles of 20 spaced 20 fishes 4 fishes fishes = 250 No. fishes taken 180 (90%) 176 (88%) 191 (96%) '0 Kleptoparasitic attempts 7 (3.9%) 13 (7.4%) 20 (10.5%) • 200

Prey dispersionand group size.--The rate of o 150 kleptoparasiticattempts on Kelp Gulls carrying fishesat Hout Bay harbor increasedas prey dis- • 100 persionincreased (X 2 = 5.5, df = 2, P < 0.01). E When 20 individually spacedfishes were pre- z sented to Kelp Gulls, the rate of kleptoparasi- 50 tism was twice as high as when the samenum- ber of fisheswere presentedin one pile (Table 4). 3 5 7 9 11 13 15 17 19 The chance of an individual kleptoparasite Number of kleptoparasites gaining food decreasedas the number of klep- involved in pusuits toparasitesin a grouppursuit increased (Fig. 1), Fig. 2. Frequency distribution of singleton and and pursuitsby a lone kleptoparasitewere more group kleptoparasiticpursuits by Kelp Gulls at study frequent than attemptsinvolving severalgulls sites. (Fig. 2). Participation in kleptoparasitic at- temptsby groupsof Kelp Gulls showedno age- related bias (Fig. 3).

0.30j

2025-

eo.2o 1• 15 • 0.15

10 o.,o •. 0.05

0.00 ...... I 3 5 7 9 11 13 15 17 9 0 Number of kleptoparasites 2 3 4 5 6 7+ Involved in pursuit Number of kleptoparasites Fig. 1. Probabilityof individual Kelp Gull klep- involved in pursuits toparasitesuccessfully gaining prey item in relation to number of kleptoparasitesinvolved in pursuit (n Fig. 3. Proportionof juvenile Kelp Gulls involved = 671). in group kleptoparasiticpursuits. 852 STEELEAND HOCKEY [Auk, Vol. 112

TABLE5. Frequencyand successof Kelp Gull kleptoparasiticactivity by age class.

Undisturbed Food-supplemented OB SMB HB SD

Adult Mean no. at site (percentof population) 20.7 (88.4) 20.6 (83.5) 150 (61.8) 403 (73.9) No. observedas kleptoparasites(percent of all kleptoparasiticattacks at site) 0 (0) 46 (48.9) 50 (29.2) 586 (60.7) Percentsuccess as kleptoparasite -- 28.3 14.0 15.7 No. observed as host (percent of hosts) 1 (100.0) 71 (89.9) 14 (38.9) 181 (69.6) Percent of hostsretain prey 100.0 71.8 57.1 60.2 Subadult Mean no. at site (percentof population) 1.9 (8.2) 2.4 (9.7) 42.3 (17.4) 73.7 (13.5) No. observedas kleptoparasites(percent of all kleptoparasiticattacks at site) 0 (0) 12 (12.8) 48 (28.1) 218 (22.6) Percentsuccess as kleptoparasite -- 33.3 20.8 10.1 No. observed as host (percent of hosts) 0 (0) 5 (6.3) 6 (16.7) 46 (17.7) Percent of hostsretain prey -- 60.0 50.0 63.0 Juvenile Mean no. at site (percent of population) 0.8 (3.4) 1.7 (6.8) 50.5 (20.8) 68.8 (12.6) No. observedas kleptoparasites(percent of all kleptoparasiticattacks at site) 1 (100) 36 (38.3) 73 (42.7) 162 (16.8) Percentsuccess as kleptoparasite 0 19.4 12.3 16.1 No. observedas host (percent of hosts) 0 (0) 3 (3.8) 16 (44.4) 33 (12.7) Percenthosts retain prey -- 33.3 68.7 63.6

Combining data from all study sites, group During food-choiceexperiments, kleptopar- pursuitson averagelasted significantly longer asitism was significantly asymmetrically dis- (8.5 + 10.2 s, n = 121) than chasesby single tributed among age classesof Kelp Gull (X 2 = gulls (4.8 + 7.9 s, n = 96; t = 2.93, df = 215, P 32.7, P < 0.01). Juveniles constituted 28.5% of < 0.01). The averagedurations of successfuland all Kelp Gulls present (n = 105 counts), but unsuccessfulpursuits were not significantlydif- accountedfor 53.3%of kleptoparasiticpursuits. ferent in either group or singleton pursuits (t- In contrast,53.5% of Kelp Gulls present were test). adults, but these accountedfor only 38.1% of Gull age.--At all sites where intraspecific kleptoparasiticattacks. kleptoparasitismamong Kelp Gullsoccurred of- Although juvenile Kelp Gulls attempted ten, the rate of attempted kleptoparasitism,by kleptoparasitismmore often than expected,we age class,was significantlydifferent from that found no differences among age classeseither expectedfrom the age structureof the popu- in their successat stealing food (X • = 1.4, ns) lation (SD, X 2 = 353.9; HB, X 2 = 26.9; SMB, or at retaining prey during a kleptoparasiticat- X 2 = 151.3;all P < 0.01; Table 5). Juvenile Kelp tack (X • = 1.2, ns; Table 5). Gulls kleptoparasitizedconspecifics more frequently than expectedgiven their abundance CONTROL OF INTRASPECIFIC KLEPTOPARASITISM in the population (SD, X • = 37.9;HB, X • = 13.1; SMB, X 2 = 9.0; all P < 0.01). Model I indicates that individual rewards for Adult Kelp Gulls never were observedat- Kelp Gulls at Strandfonteindump employing tempting intraspecifickleptoparasitism at Oli- a mixed foraging strategy are greatestwhen a fantsbosBay and, at all other sites, adults at- small proportion of the available time is allo- tempted intraspecifickleptoparasitism signifi- cated to kleptoparasitism(Fig. 4). The stable cantly lessoften than expected(SD, X • = 311.2; compositionfor flocksof Kelp Gulls at Strand- HB, X • = 37.6; SMB, X 2 = 16.1; all P < 0.01). fontein dump comprisingindividuals follow- Subadultsgenerally were involved in klepto- ing one of two foraging strategies,kleptopar- parasiticincidents in proportion to their rela- asitismand scavenging(Model II), is calculated tive abundancein the population(Table 5). in Table 6. Although simplistic,this calculation October1995] KleptoparasitismAmong Kelp Gulls 853

8O specifickleptoparasitism among Kelp Gulls.Prey cannot be stolen if their handling time is less 70- than the time neededfor a kleptoparasiticattack (Barnardand Thompson1985). At Olifantsbos 60- Bay, where Kelp Gulls fed on small inverte- brateswith handling timesof lessthan 1 s, klep- toparasitismrarely was possible. Large prey take 50- longer to handle than small prey, and gulls preferentiallyselected small, rapidly handled 40- prey that had a concomitantlylow probability Ii'"i!111 of being stolen (Table 2). 30- I/ Studiesof a range of kleptoparasiticinterac- I I tions have shown that increasingprey size and 20 ß handling time increaseboth the rate (Hopkins 'lI and Wiley 1972,Fuchs 1977, Fischer 1985, Hackl 10 ß and Burger1988, Hockey and Steele1990, Bur- ger and Gochfeld1991) and success(Gochfeld 0 • and Burger1981, Barnard and Thompson1985, 0.01 0.2 0.4 0.6 0.8 0.99 Hackl and Burger1988, Hockey and Steele1990) of kleptoparasitism.However, both Dunn (1973) Proportion of time and Fuchs(1977) found that kleptoparasiticsuc- spent kleptoparasitizing cessamong terns (Sterna spp.) was inverselyre- Fig. 4. Averageequivalent number of food items lated to prey size, and attributed this to in- consumedper hour (individual reward)by individual creasedhost vigilance when carryinglarge prey. Kelp Gulls at Strandfonteindump for different klep- Prey size was positively related to kleptopar- toparasitismfrequencies. asiticsuccess among Kelp Gulls in the southwestern Cape. While the rate of kleptoparasitismmay de- demonstratesthat in any flock the number of creaseas food availability increases (Dunn 1973), individuals obtaining food solely through in- it has been suggestedthat hosts may give up traspecifickleptoparasitism represents a very food more readily when food is abundant(Birt small proportionof that flock. and Cairns 1987).This is supportedby the re- For example,in a flock of 13 gulls where all suits of our study, where high kleptoparasitic individuals are scavengersand do not resort to success rates were recorded at the two sites kleptoparasitism,each gull will gain the equiv- where food availability was high (Table 1). Al- alent of 55.2 average-sizedfood items per hour. though the successof kleptoparasiticattempts If one of these gulls changed foraging strategy at 16 Mile Beach(where large Donaxwere cap- and becamea kleptoparasite,it would improve tured infrequently) was even higher (Table 1), its hourly reward to the equivalent of 58.0 av- almostcertainly this was a consequenceof the erage-sized food items. However, if a second prey-handlingtechnique, where the gulls lose gull alsobecame a kleptoparasite,the two klep- direct contactwith their prey when musselsare toparasiteswould eachonly gain the equivalent droppedto break them open. The risksto the of 26.6 average-sizedfood items per hour. As hostattending this prey-handlingtechnique are this is less than the hourly reward of a scav- directly proportional to the number of times enging gull, it would not benefit the second the shell has to be dropped before it breaks gull to becomea kleptoparasite.Therefore, a (Hockey and Steele 1990).Kleptoparasitism un- flock of 13 gulls would be stable when 12 of der theseconditions is a very different behavior the gulls are scavengersand only 1 a klepto- than when the parasiteis "chasing"or "food parasite. snatching."However, certain preconditionsap- DISCUSSION ply to all of thesebehaviors, in particularthe availabilityof hostsand the opportunityto the Factorsinfluencing rate andsuccess of kleptopar- kleptoparasites.Thus, we havetreated all forms asitism.--Preycharacteristics are important in of intraspecifickleptoparasitism as falling along determining both the rate and successof intra- a behavioral continuum. 854 STEELEAND HOCKEY [Auk, Vol. 112

TABLE6. Calculated average reward for individual kleptoparasites(in terms of number of equivalent mean- sizedfood itemsconsumed per hour) in different-sizedflocks of foraging Kelp Gulls at Strandfonteindump, comprisingscavengers and kleptoparasites.Individual rewardfor scavengersis 55.2(see Appendix 2). Stable compositionfor each flock (i.e. the point at which no individual can improve its rate of food consumption by changingforaging strategy) is given in bold. Flock Numberof kleptoparasites size 1 2 3 4 5 6 7 8 9 10

2 4.8 3 9.7 2.4 4 14.5 4.8 1.6 5 19.3 7.2 3.2 1.2 6 24.2 9.7 4.8 2.4 1.0 7 29.0 12.1 6.4 3.6 1.9 0.8 8 33.8 14.5 8.1 4.8 2.9 1.6 0.7 9 38.6 16.9 9.7 6.0 3.9 2.4 1.4 0.6 10 43.5 19.3 11.3 7.2 4.8 3.2 2.1 1.2 0.5 11 48.3 21.7 12.9 8.5 5.8 4.0 2.8 1.8 1.1 0.5 12 53.2 24.2 14.5 9.7 6.8 4.8 3.5 2.4 1.6 1.0 13 58.0 26.6 16.1 10.9 7.7 5.6 4.1 3.0 2.1 1.4 14 62.8 29.0 17.7 12.1 8.7 6.4 4.8 3.6 2.7 1.9 15 67.6 31.4 19.3 13.3 9.7 7.2 5.5 4.2 3.2 2.4 16 72.5 33.8 20.9 14.5 10.6 8.1 6.2 4.8 3.8 2.9 17 77.3 36.2 22.5 15.7 11.6 8.9 6.9 5.4 4.3 • 3.4 18 82.1 38.6 24.2 16.9 12.6 9.7 7.6 6.0 4.8 3.9 19 86.9 41.1 25.8 18.1 13.5 10.5 8.3 6.6 5.4 4.3 20 91.8 43.5 27.4 19.3 14.5 11.3 9.0 7.2 5.9 4.8 25 115.9 55.5 35.4 25.4 19.3 15.3 12.4 10.3 8.6 7.2 30 140.1 67.6 43.5 31.4 24.2 19.3 15.9 13.3 11.3 9.7 35 164.2 79.7 51.5 37.4 29.0 23.3 19.3 16.3 14.0 12.1 40 188.4 91.8 59.6 43.5 33.8 27.4 22.8 19.3 16.6 14.5 45 212.5 103.8 67.6 49.5 38.6 31.4 26.2 22.3 19.3 16.9 50 236.7 115.9 75.7 55.5 43.5 35.4 29.7 25.4 22.0 19.3 55 260.8 128.0 83.7 61.6 48.3 39.4 33.1 28.4 24.7 21.7 60 285.0 140.1 91.8 67.6 53.1 43.5 36.6 31.4 27.4 24.2 65 309.1 152.1 99.8 73.7 58.0 47.5 40.0 34.4 30.1 26.6 70 333.3 164.2 107.9 79.7 62.8 51.5 43.5 37.4 32.7 29.0 75 357.4 176.3 115.9 85.7 67.6 55.5 46.9 40.5 35.4 31.4 80 381.6 188.4 124.0 91.8 72.5 59.6 50.4 43.5 38.1 33.8 85 405.7 200.4 132.0 97.8 77.3 63.6 53.8 46.5 40.8 36.2 90 429.9 212.5 140.1 103.8 82.1 67.6 57.3 49.5 43.5 38.6 95 454.0 224.6 148.1 109.9 86.9 71.6 60.7 52.5 46.2 41.1 100 478.2 236.7 156.2 115.9 91.8 75.7 64.2 55.5 48.8 43.5

An increase in the rate of kleptoparasitism Krebs and Barnard (1980) suggestedthat the with increasingspacing of prey (Table 4) ap- rate of kleptoparasitism should increase with pearsinitially to be counterintuitive. However, population density. Barash et al. (1975) found Kelp Gulls apparentlywere better able to judge that gulls modified their prey-handlingtech- the amount of available food remaining when nique as gull density increased,apparently to prey were dispersed. As prey abundance de- limit prey lossthrough kleptoparasitism.In our creased,newly arrived gulls resorted to klep- study the rate of kleptoparasitismwas highest toparasitismin order to obtain some of the few at 16 Mile Beach,where gull densitywas lowest remaining prey. When fisheswere presentedin (Table 1); however, we did not set out specifi- a single pile, a dense crowd of fighting gulls cally to test this relationship.While the density immediately formed over the food and ap- of gulls foraging at a site is likely to have a proaching gulls joined this crowd, which per- marked effect on the rate of kleptoparasitism sisted for some time after all the food was taken, within a population, prey abundanceand han- ratherthan attemptingto kleptoparasitizebirds dling time probablyare of greaterimportance. leaving the area with fish. Prey lossby hoststhrough kleptoparasitism October1995] KleptoparasitismAmong Kelp Gulls 855

has been correlatedwith the number of klep- Although intraspecifickleptoparasitism is a toparasitesinvolved in the pursuit (Hatch 1970, functionalparallel of the interspecificbehavior 1975, Barnard and Thompson 1985, Hackl and (Brockmann and Barnard 1979), it presents a Burger 1988), although the probability of an more complex situation. Intraspecificklepto- individual kleptoparasitegaining food decreas- parasitismmay be an efficient foraging tech- es as the number of pursuersincreases (Hatch nique for some speciesin some situations,but 1970,1975; Fig. 1),accounting for the relatively not all individuals within a flock can resort to low frequencyof grouppursuits (Fig. 2). intraspecifickleptoparasitism, except at very low Our findings support conclusionsof several frequencies.Obviously, some individuals with- prior studies of both inter- and intraspecific in the populationmust hunt in order to provide kleptoparasitismthat a suite of factorsinfluence the resourcebase on which the kleptoparasitic the rate and successof this behavior. These fac- behavior depends. tors include characteristicsof both prey items The resultsof Model II give an indicationof and the gull population.Although the rate of how few obligate intraspecifickleptoparasites kleptoparasitismat most sites generally is low, a KelpGull flockcan support. The averagegroup the risk of food lossthrough kleptoparasitism of 30 Kelp Gulls activelyforaging at Strandfon- is sufficientlygreat to influencefood choice by tein dump would be able to supporta maximum gulls. of only 2 obligate kleptoparasites(Table 6). Mechanismscontrolling the rate of intraspecificHowever, it was clear from field observations kleptoparasitism.--Therehas been debate re- that more than two individuals were respon- gardingthe efficacyof kleptoparasitismas a for- sible for the kleptoparasitizingof conspecifics. agingtechnique, and it hasbeen suggested that Field observationsand experimental studies kleptoparasitism is a less efficient means of of age-related kleptoparasitismamong Kelp gainingfood than is hunting (e.g.Furness 1977, Gulls indicate that the frequencywith which Kushlan1978, 1979, LeBaron and Heppner 1985). this techniqueis usedto obtain prey decreases However, Ens et al. (1990) have shown that, for with ageand, thus,is not an individually fixed, EurasianCurlews (Numeniusarquata) on mud- obligatetrait. This is supportedby the empir- flats in the Netherlands,kleptoparasitism is a ically basedmodels: the stableratio of obligate profitableform of foraging,despite the low pro- scroungersto obligateproducers predicted by portion of success.Similarly, Dunbrack (1979) Model II is too low to accord with the number concludedthat, contrary to Kushlan's(1978) of birdsobserved kleptoparasitizing in the field. findings, kleptoparasitismwas profitable in Among Bald Eagles(Haliaeetus leucocephalus), termsof prey returnsfor GreatEgrets (Casmer- populationwide frequencies of intraspecific odiusalbus) foraging in Florida. kleptoparasitismand hunting are balancedat The resultsgenerated by Model I indicatethat the evolutionarily stable strategy (ESS) point kleptoparasitismcan be an efficient method of (sensuMaynard-Smith and Parker 1976), such foraging, and that the inclusion of a low rate that the rewardsaccrued through each strategy of kleptoparasitismin a gull's foragingreper- are nearly equal (Hansen 1986). This situation toire can raise its food intake rate above that is the evolutionarilystable population strategy achieved by scavengingalone. Basedon ma- describedby Treisman(1977). Similarly, the po- nipulation of parametervalues in Model I (see pulationwide rate of intraspecifickleptopara- Appendix1), the key conditionproducing this sitism that maximizes the rate of food intake resultis that the prey itemsobtained through for Kelp Gullsforaging at Strandfonteindump, kleptoparasitismmust, on average,be larger as shown in Model I, would be an evolution- than thoseobtained by scavenging.If individ- arily stablepopulation strategy. Every gull in uals were to obtain the same-sized food items the flock would not have to allocate 15% of its throughkleptoparasitism and scavenging, then available foraging time to kleptoparasitismin kleptoparasitismwould alwaysreduce the in- order to maximize individual rate of food intake dividual reward to the kleptoparasiteand, thus, (Fig. 4), but the overall averagefood intake rate wouldnot be a viableforaging technique. How- would be maximizedwhen the populationwide ever, becausesuccessful kleptoparasites gain average percent of available time allotted to larger than average-sizedfood items,low rates kleptoparasitismwas about 15%. of kleptoparasitismcan enhancethe rate of food Studies of several gull specieshave shown intake. that juveniles are lessefficient at hunting than 856 STEELEAND HOCKE• [Auk,Vol. 112

adultsat both natural (e.g. Searcy1978, Greig Monaghan 1980),but whether generalizations et al. 1983,Maron 1983,Hockey et al. 1989)and canbe madebetween species is debatable.Adult supplementedfeeding sites (e.g. Verbeek 1977a, Herring Gulls(L. argentatus)in northeasternEn- 1977b,Maclean 1986). In many cases,juvenile gland usesupplementary feeding sites,such as gulls also have been shown to be less efficient dumps, more often than juveniles. These ap- at kleptoparasitismthan adults (Burger and parently are preferred feeding sites for adult Gochfeld 1979, 1981, Carrol and Cramer 1985, males,which dominateadult femalesand ju- Hesp and Barnard 1989, Hockey et al. 1989). veniles,and areable to relegatethe latter to less JuvenileKelp Gullsin our study,however, were preferred sites(Monaghan 1980).Among Kelp asefficient at kleptoparasitism(both in effecting Gulls in both South Africa and Chile, it is ju- it and avoidingit) aswere adults;similar results venilesthat aggregatedisproportionately at such have been reported by Verbeek (1977c) and feeding sites (Hockey and Steele 1990). This Hackl and Burger(1988). Thus, it is predictable suggestseither that higher-quality food was that juvenile Kelp Gullsshould use intraspecific available at the English refuse dumps than at kleptoparasitismin order to exploit the greater those studied in South Africa and Chile, or that hunting capabilitiesof the adults, and should dominance hierarchies in the two speciesop- resortto kleptoparasitismmore often than adults erate differently. (seealso Pettitt 1953,Burger and Gochfeld1979). Kelp Gulls kleptoparasitizeinterspecifically, Juvenilegulls beg from, and are fed by, their as well as intraspecifically,stealing food from parentsfor a short period after fledging (e.g. a variety of species,including other gulls and Burger 1981, Holley 1982). Intraspecific klep- African Black Oystercatchers(H. moquini;e.g. toparasiticbehavior may develop from thisbeg- Hockey 1980). While the physicalact of food ging behavior, particularly at the time when stealingmight be similar, independent of the adults stop feeding the juveniles. host species,inter- and intraspecifickleptopar- The observationthat inefficient hunters (ju- asitismmay not have the sameevolutionary or- veniles)rely moreheavily on intraspecificklep- igins. Although the opportunitiesfor klepto- toparasitismthan experienced, older birds has parasitismalways are limited by host availabil- a counterintuitive corollary that older birds do ity, intraspecific kleptoparasitism is further not exerttheir presumeddominance by stealing constrained by an additional feedback loop. from young birds. Among EurasianOyster- Time spent in kleptoparasitismcan be equated catchers(Haematopus ostralegus), the amount of with lost hunting time, thereby reducing the food gained by kleptoparasitismincreases with overall availability of prey that can be stolen. dominance status (Ens and Goss-Custard1984). Thus, it is incorrectto view intraspecificklep- However, young oystercatcherswhen they first toparasitismas simply an extension of the in- arrive on the nonbreedinggrounds obtain more terspecificbehavior. of their food by kleptoparasitism(directed both at other juveniles and adults) than do adults ACKNOWLEDGMENTS (Gross-Custard and dit Durell 1987). The frequencyof this behaviorsubsequently decreases We thank the CapeTown City Councilfor permis- because of the establishment of dominance hi- sion to work at the Strandfontein (CoastalPark) refuse erarchies,and remains low for several years dump.Financial assistance from the Universityof Cape (Goss-Custardand dit Durel11987).Comparable Town's Research Committee, the Foundation for Re- effectsof dominanceare not evidentamong Kelp searchDevelopment, and the South African National Gulls; older birds initiate fewer attacks than Committeefor OceanographicResearch is gratefully acknowledged. youngerbirds (Hockey and Steele1990, current study). However, a larger than expectedpro- portion of all attacksare directedat juveniles LITERATURE CITED and subadults.This suggeststhat if dominance BARASH, D. P., P. DONOV^N, AND R. MYPdCK. 1975. influences kleptoparasiticbehavior in Kelp Clam dropping behaviorof the Glaucous-winged Gulls, as clearly is the casein EurasianOyster- Gull (Larusglaucescens). Wilson Bull. 87:60-64. catchers,it functionsthrough dominant birds B^RNXRD,C. J., AND R. M. SIBL¾. 1981. Producers and being targetedless often for attackrather than scroungers:A generalmodel and its application using their dominanceto initiate attacks. to feeding flocksof House Sparrows.Anim. Be- Dominancehierarchies exist in gulls (e.g. hay. 29:543-550. October1995] KleptoparasitismAmong Kelp Gulls 857

BARNARD,C. J., ANDD. B. A. THOMPSON.1985. Klep- ferencesin piracyof frigatebirdsfrom Laughing toparasitism:Host and prey selectionby gulls. Gulls. Condor 83:79-82. Pages217-254 in Gulls and plovers:The ecology GOSs-CUSTARD,J. D., AND S. E. A. LE V. DIT DURELL. and behaviour of mixed-speciesfeeding groups 1987. Age-relatedeffects on Oystercatchers,Hae- (C. J. Barnardand D. B. A. Thompson,Eds.). Croom matopusostralegus, feeding on mussels,Mytilus Helm, London. edulis,II. Aggression.J. Anim. Ecol.56:537-548. BIRT,V. L., ANDD. K. CAIRNS. 1987. Kleptoparasitic GREIG,S. A., J. C. COULSON,AND P. MONAGHAN. 1983. interactionsof Arctic SkuasStercorarius parasiticus Age-relateddifferences in foragingsuccess in the and BlackGuillemots Cepphus grylle in north-east- Herring Gull Larusargentatus. Anim. Behav.31: ern Hudson Bay,Canada. Ibis 129:190-196. 1237-1243. BROCKMANN,H. J., AND C. J. BARNARD.1979. Klep- HACKL,E., AND J. BURGER.1988. Factorsaffecting toparasitismin birds. Anim. Behav.27:487-514. piracyin Herring Gullsat a New Jerseylandfill. BURGER,J. 1981. Feeding competition between Wilson Bull. 100:424-430. Laughing Gulls and Herring Gulls at a sanitary HANSEN,A.J. 1986. Fighting behavior in Bald Ea- landfill. Condor 83:328-335. gles:A test of gametheory. Ecology67:787-797. BURGER,J. 1981. Feeding competition between HARRISON,P. 1983. Seabirds:An identification guide. Laughing Gulls and Herring Gulls at a sanitary Croom Helm, Beckenham,England. landfill. Condor 83:328-335. HATCH,J.J. 1970. Predationand piracy by gulls at BURGER,J., AND M. GOCHFELD.1981. Age-relateddif- a ternery in Maine. Auk 87:244-254. ferencesin piracy behaviourof four speciesof HATCH,J. J. 1975. Piracy by Laughing Gulls Larus gull (Larusspp.). Behaviour 77:242-267. atricilla:An exampleof the selfishgroup. Ibis 117: BURGER,J., AND M. GOCHFELD.1991. The Common 357-365. Tern: Its breedingbiology and socialbehaviour. HESP,L. S., ANDC. J. BARNARD.1989. Gulls and plov- Columbia Univ. Press, New York. ers: Age-relateddifferences in kleptoparasitism CARROL,S. P., ANDK. L. CRAMER.1985. Age differ- among Black-headedGulls (Larusridibundus). Be- encesin kleptoparasitismby LaughingGulls (Lar- hay. Ecol. Sociobiol. 24:297-304. usatricilla) on adult and juvenile Brown Pelicans HOCKEY,P. A. R. 1980. Kleptoparasitismby Kelp (Pelecanusoccidentalis). Anim. Behav. 33:201-205. Gulls Larusdominicanus of African Black Oyster- CRAWFORD,R. J. M., J. COOPER,AND P. A. SHELTON. catchersHaematopus moquini. Cormorant 8:97-98. 1982. Distribution, population size, breedingand HOCKEY, P. A. R., A. L. BOSMAN,AND P. G. RYAN. 1989. conservationof the Kelp Gull in southernAfrica. Age-related intraspecific kleptoparasitism and Ostrich 53:164-177. foraging successof Kelp Gulls Larusdominicanus. DUNBRACK,R. L. 1979. A re-examinationof robbing Ardea 77:205-210. behavior in foraging egrets.Ecology 60:644-645. HOCKEY,P. A. R., AND W. K. STEELE.1990. Intraspe- DUNN,E.K. 1973. Robbingbehavior of RoseateTerns. cifickleptoparasitism and foragingability ascon- Auk 90:641-651. straintson food selectionby Kelp Gulls Larus ENS,B. J., P. ESSELINK,AND L. ZWARTS. 1990. Klep- dominicanus.Pages 679-706 in Behaviouralmech- toparasitismas a problem of prey choice:A study anisms of food selection (R. N. Hughes, Ed.). on mudflat-feeding Curlews Numeniusarquata. Springer-Verlag,Berlin. Anim. Behav. 39:219-230. HOLLEY,A. J.F. 1982. Post-fledginginteractions on ENS, B. J., AND J. D. GOss-CusTARD.1984. Interfer- the territory betweenparents and young Herring ence among Oystercatchers,Haematopus ostrale- Gulls Larusargentatus. Ibis 124:198-203. gus,feeding on mussels,Mytilus edulis, on the Exe HOPtaNS,C. D., ANY R. H. WILEY. 1972. Food para- estuary.J. Anim. Ecol.53:217-231. sitismand competitionin two terns.Auk 89:583- FISCHER,D. L. 1985. Piracy behavior of wintering 594. Bald Eagles.Condor 87:246-251. KINSKY, F.C. 1963. The Southern Black-backed Gull FUCHS,E. 1977. Kleptoparasitismof SandwichTerns (Larus dominicanus)Lichtenstein, measurements, by Black-headedGulls. Ibis 119:183-190. plumagecolour and moult cycle.Rec. Dom. Mus. FURNESS,R. W. 1977. Great Skuas as predators of (Wellington) 4:149-219. mammals. Scottish Birds 9:319-321. KREBS,J. R., AND C. J. BARNARD. 1980. Comments on FURNESS,R. W. 1987. Kleptoparasitism in seabirds. the function of flocking in birds. Pages795-799 Pages77-100 in Seabirds:Feeding ecology and in Acta XVII CongressusInternationalis Orni- role in marine ecosystems(J.P. Croxall, Ed.). thologici (R. N6hring, Ed.). Berlin, 1978.Deutsche CambridgeUniv. Press,Cambridge. Ornithologen-Gesellschaft,Berlin. GIRALDEAU,L.-A., ANDL. LEFEBvRE.1986. Exchange- KUSHLAN,J. 1978. Nonrigorousforaging by robbing able producer and scroungerroles in a captive egrets.Ecology 60:649-653. flock of feral pigeons:A casefor the skill pool KUSHLAN,J. 1979. Short-termenergy maximization effects. Anim. Behav. 34:797-803. of egret foraging.Ecology 60:645-646. GOCHFELD,M., ANDJ. BURGER.1981. Age-related dif- LEBARON,G. S., AND F. H. HEPPNER. 1985. Food theft 858 STEELEAND HOCKEY [Auk, Vol. 112

in the presenceof abundantfood in Herring Gulls. STEELE,W. K. 1992. Diet of Hartlaub's Gull Larus Condor 87:430-431. hartlaubiiand the Kelp Gull L. dominicanusin the MACLEAN,A. A. E. 1986. Age-specificforaging abil- southwestern Cape Province, South Africa. Os- ity and the evolution of deferred breeding in trich 63:68-82. three speciesof gulls. Wilson Bull. 98:267-279. STEELE,W. K., ANDP. A. R. HocKEY. 1990. Population MACLEAN, G.L. 1993. Roberts' birds of southern Af- size, distributionand dispersalof Kelp Gulls in rica. John Voelcker Bird BookFund, Cape Town. the southwesternCape, South Africa. Ostrich 61: MARON,J.L. 1983. Shell-droppingbehaviorofWest- 97-106. ern Gulls (Larusoccidentalis). Auk 99:565-572. TREISMAN,g. 1977. The evolutionary restrictionof MAYNARD-SMITH,J., AND G. A. PARKER. 1976. The aggressionwithin a species:A gametheory anal- logic of asymmetriccontests. Anim. Behav. 24: ysis. J. Math. Psychol. 16:167-203. 159-175. VERBEEK,N. A. g. 1977a. Age differences in the MONAGHAN,P. 1980. Dominance and dispersal be- diggingfrequencies of Herring Gullson a dump. tween feeding sites in the Herring Gull (Larus Condor 79:123-125. argentatus).Anim. Behav.28:521-527. VERBEEK,N. A.M. 1977b. Comparativefeeding be- NATIONAL RESEARCH INSTITUTE FOR NUTRITIONAL DI- havior of immatureand adult Herring Gulls. Wil- SEASES(SOUTH AFRICA) (N.R.I.N.D.). 1986. Food son Bull. 89:415-421. compositiontables. South African Medical Re- VERSEEK,N. A.M. 1977c. Interactions between Her- search Council, Parow. ring and LesserBlack-backed gulls feeding on PErrift, R.G. 1953. Comparativeaggressiveness of refuse. Auk 94:726-735. the first-year and adult Black-headedGulls. Br. VICKERY,W. L., L.-A. GIRALDEAU,J. J. TEMPLETON,D. Birds 45:333-334. L. KRAMER, AND C. A. CHAPMAN. 1991. Produc- SEARCY,W. A. 1978. Foraging successin three age ers,scroungers and groupforaging. Am. Nat. 137: classesof Glaucous-wingedGulls. Auk 95:586- 847-863. 590.

APPENDIX1. Parameters,assumptions, and calculationsused to model individual reward for Kelp Gulls at Strandfonteindump, at varying ratesof kleptoparasitism.Individual reward expressedas numberof food items, standardizedto averagesize consumedper hour.

Parameters.--Meanfood-handling time (TH) = 3.8 + 11.9 s (n = 40); mean interval between food items during scavenging(including TH) = 59.8 + 47.8s (n = 133).Therefore, mean food intake rate during scavenging (FR) = 60 items/h. Mean kleptoparasiticpursuit time (TP) = 8.3 + 11.4 s (n = 138). Mean food length during scavenging= 32 + 27 mm (n = 165);mean length of food items successfullykleptoparasitized = 75 + 52 mm (n = 68). Therefore,mean length of food itemslost to kleptoparasitism,relative to meanfood sizeand expressed as a proportion(K) = 2.3. Proportionof successfulkleptoparasites during both group and singletonpursuits (KS) = 0.14. Proportionof availableforaging time allocatedto scavenging(Q) = variable (between1.00 and 0.01). The proportionof availableforaging time allocatedto kleptoparasitism(P) = variable (between0 and 0.99). Assumptions.--(1)Number of Kelp Gulls in flock able to interactwith each other (N) = 100;(2) 25% of all food items found during scavengingare large enoughto warrant kleptoparasiticattempt; (3) all Kelp Gulls in flock follow sameforaging "strategy"and kleptoparasitizeat same rate. Calculations.--Value1: Food items found during scavenging= (QM.FR). Value 2: Food items lost to kleptoparasites= (Q.FR.O.25.KS.K).Value 3: Fooditems consumedby individual during scavenging= (Value 1 - Value 2). Value 4: Maximum number of kleptoparasiticpursuits possible per hour = P/[TP + (KS.TH)]. Value 5: Number of food items available to be kleptoparasitized= (N.Q.FR.O.25)/(N.P). Value 6: Actual numberof kleptoparasiticpursuits possible = Value 5 up to, but not exceeding,Value 4. Value 7: Number of successfulkleptoparasitic attempts = Value 6.KS. Total food intake = (Value 7.K) + Value 3. October1995] KleptoparasitismAmong Kelp Gulls 859

APPENDIX2. Parameters,assumptions, and calculationsused to model stable compositionof different sized flocksof foraging Kelp Gulls at Strandfonteindump, which are comprisedof obligate scavengersand obligatekleptoparasites. Stable point is that at which no individual can improveits rate of food intake per hour by changing foraging strategy.

Parameters.--Meaninterval between food items during scavenging(including food handling time) = 59.8 + 47.8 s (n = 133).Therefore, mean food intake rate during scavenging(FR) = 60 items/h. Mean food length during scavenging= 32 + 27 mm (n = 165);mean length of food itemssuccessfully kleptoparasitized = 75 + 52 mm (n = 68). Therefore,mean length of food itemslost to kleptoparasitismrelative to meanfood size expressedas a proportion(K) = 2.3. Proportionof successfulkleptoparasites during both groupand singleton pursuits(KS) = 0.14;number of kleptoparasitesin flock(P) = variable(between 1 and 10);number of scavengers in flock (Q) = variable (between 1 and 99). Assumption.--Ofall food itemsfound during scavenging,25% are large enoughto warranta kleptoparasitic attempt. Calculations.--Individualkleptoparasite reward (Y) = (Q. FR.0.25.KS.K)/P; individual scavengerreward = [(Q.FR) - (Y.P)]/Q. Thus,individual scavengers gain equivalent of 55.2mean-sized food items. Therefore, unless kleptoparasites gain more than this amount it would benefit them to changestrategy to becomescavengers.