Anim. Behav., 1997,53, 197-208

Nest-site

selection in Savannah sparrows: using gulls as scarecrows?

NATHANIEL T. WHEELWRIGHT, JOSHUA J. LAWLER& JOSHUA H. WEINSTEIN

Department of Biology, Bowdoin College

Received 8 May 1995; initial acceptance 2 September 1995;final acceptance 20 May 1996; MS. number: A7324R)

Abstract. Savannah sparrows, Passercu/us sandwichensis, breeding on Kent Island, New Brunswick,Canada, have two types of nest predators, one of them (herring gulls, LaTUS argentatus) abundant butrelatively ineffective, the other (American crows, Corvus brachyrhynchos) scarce but highly effective. Wehypothesized that the net effect for Savannah sparrows of nesting near gulls would be to reduce theoverall risk of nest predation. Despite being surrounded by predators, the eggs and offspring ofsparrows that nested among gulls survived as well during the incubation and post-fledging periods asdid those of sparrows that did not nest among gulls. During the nestling period, sparrows nestingamong gulls had significantly lower predation rates. In defending their own nests from predatory crows,gulls apparently shielded nearby sparrows from the more dangerous predator. Experiments with modelpredators demonstrated that sparrows reacted to gulls as potential predators of their eggs and nestlings.Sparrows apparently recognized crows as a far greater threat, however. The tendency to nest near gullsappeared not to be heritable or influenced by early experience. Sparrows nes~g among gulls wereindistinguishable from sparrows nesting away from gulls in terms of body size, age and date of nesting.By choosing nest sites in microhabitats that gulls avoided, such as dense patches of goldenrod andblueberry, and by adopting more cautious approaches to their nests, sparrows nesting near gullsreduced their risk of predation by gulls. The density of Savannah sparrow nests was inversely correlatedwith the density of gull nests, which suggests that sparrows avoided gulls despite the apparent advantagein terms of reduced nest predation by crows. A strong nesting association between gulls and birds likeSavannah sparrows is unlikely to evolve because of the low heritability of the trait, gene flow from otherpopulations where avoiding gulls and other potential predators is selectively advantageous, andconstraints on short-lived birds in learning to differentiate situations in which a predator presents athreat from those in which it provides protection. @ 1997 The Association for the Study of Animal Behaviour

Birds often locate their nests near predatorsdespite what seem to be substantial risks to them-selves or to their offspring. Researchers haveoffered two explanations for why birds nest nearpredators. The first assumes that nesting nearpredators is risky but emphasizes that birds areconstrained in acquiring safer nest sites. Eventhough nesting near predators may reduce repro-ductive success compared with nesting elsewhere,birds that are young, ageing, ailing or sociallysubordinate may have few other options {Blus &Keahey 1978; Reese & Kadlec 1985; Lessels &

Correspondence: N. T. Wheelwright, Department ofBiology, Bowdoin College, Brunswick, ME 040 11,U.S.A (email: bss@bowdoin.edu). J. Lawler is nowat the Fish and Wildlife Department, Utah StateUniversity, Logan, UT 84322-5210, U.S.A.

Krebs 1989). Species with specialized nestingrequirem~nts, such as birds that must breed onislands that happen to be inhabited by predators,may face similar limitations (Bourget 1973). Nestassociations with predators may also be an inci-dental consequence of shared habitat preferences(Erwin et al. 1981). Inexperienced birds may un-intentionally nest near predators because they failto recognize the dangers (Wheelwright & Schultz1994). Finally, nest-site selection may hot be asflexible

a trait as is often presumed if nestingbehaviour is learned at a young age or has agenetic basis. In other words, birds raised nearpredators

may n~st near predators because (jftheir early experience or genes.

An alternative explanation for nesting nearpredators is that the benefits outweigh the greater

0003-3472/97/010197+ 12 $25.00/0/ar960288 (Q 1997 The Association for the Study of Animal Behaviour

197

Animal Behaviour, 53, 1198

risks of predation. In theory, the increased risk ofegg or nestling loss incurred by nesting nearpredators could be compensated by improvedadult survival or reduced competition for food. Incertain cases, one predator may actually provideprotection against another, more effective pred-ator. For example, common eiders, Somateriamollissima, nesting within gull .(Larus spp.) col-onies experience lower rates of nest. predationthan eiders nesting elsewhere, even though gullssometimes eat eider eggs and chicks, because gullsalso alert against or drive off more dangerouspredators such as bald eagles, Haliaeetus leuco-cephalus, or minks, Mustela vison (Bourget 1973;Gerell 1985; Gotmark & Ahlund 1986; N.T.Wheelwright, personal observation; see alsoDwernychuk & Boag 1972). Likewise, grebes(Podiceps occipitalis, Rollandia rolland) nesting inassociation with brown-headed gulls, L. maculi-pennis, may lose their nests to gulls (Burger 1984).They have higher reproductive success and loweradult mortality than grebes nesting outside gullcolonies, however, because, forewarned by thegulls about the approach of predators, the grebescan adopt appropriate anti-predator behaviours.Spotted sandpipers, Actitis macularia, nestingwithin common tern, Sterna hirundo, coloniesexperience more predation by migratory ruddyturnstones, Arenaria interpres, than do sandpipersnesting outside tern colonies, but experience lesspredation by minks, which pose the larger threatto sandpipers (Alberico etal. 1991). Birds nestingnear colonies of biting ants or stinging wasps maygain protection against predatory snakes and pri-mates (Young et al. 1990; Joyce 1993). A pre-requisite for using one predator as protectionagainst another is the ability to distinguishbetween different predators and to assess therelative danger presented by each, which is knownin a variety of bird species (Nice & ter Pelkwyk1941; Curio 1975; Walters 1990; Winkler 1992).

Although protective nest associations betweenbird species have long been documented, andplausible hypotheses have been advanced toexplain them as adaptations .(Koskimies 1957;Cullen 1960; Kruuk 1964; Drycz et a1. 1981), fewstudies have been able to discriminate between the~constraint' and 'benefit' hypotheses, because thefitness consequences (e.g. lifetime reproductivesuccess, offspring recruitment) of nesting nearversus far from predators are notoriously difficultto measure completely, information about a bird's

early experience and the heritability of nestingbehaviour is generally lacking, and studies havebeen relatively short-term or non-experimental(but see Nuechterlein 1981). We were able toavoid some of these shortcomings by takingadvantage of a marked, known-age bird popula-tion in which reproductive success and the herit-ability of behaviour could be estimated over an

8-year period.To evaluate possible advantages and disadvan-

tages of nesting near predators, we compared thebehaviour and reproductive success of Savannahsparrows, Passerculus sandwichensis, nesting in thepresenc(( and absence of herring gulls, L. argenta-ius. Although gulls prey on Savannah sparrowsand their offspring (Wheelwright & Rising 1993),we hypothesized that for Savannah sparrows thenet effect of the presence of gulls would be toreduce the overall risk of nest predation, becausethe gulls ward off more effective predators,American crows, Corvus brachyrhynchos. Weattempted to distinguish between the constrainthypothesis and the benefit hypothesis by (1) quan-tifying the relationship between nest-site selectionand fledging success, fledgling size, offspring-recruitment rate and adult survivorship; (2)determining whether the decision to nest amongpredators was related to a bird's sex, body size,age, parentage or experience as a nestling; (3)assessing how birds nesting among predatorsmight change their behaviour patterns to reducetheir risk of predation; (4) testing whether thebehaviour patterns of birds towards differentpredators were gauged in some measure to thethreats posed by the predators. We also examinedseasonal variation in predation risk and comparedthe responses of Savannah sparrows to modelpredators presented at different stages of repro-duction (Patterson et al. 1980; Knight & Temple

1986).

HypothesesOur first hypothesis was that sparrows would

recognize both gulls and crows as potential nestpredators and would distinguish them fromsimilar-sized animals that posed no threat to eggsor nestlings. We also expected the intensity of thesparrows' responses to reflect the magnitude ofthe perceived danger to their offspring. Thus,sparrows should demonstrate greater alarm inresponse to the presence of a crow near their nestthan to the presence of a gull, because crows were

Wheelwright et al.: Sparrow nest-site selection 199

observed to be more systematic and effectivenest predators. Second, we predicted that thereproductive success of Savannah sparrows nest-ing near gulls would be as high as or higher thanthat of sparrows nesting away from gulls because,although gulls prey opportunistically on sparrownests, they also tend to chase away crows, themore dangerous predator. Assuming that the firsttwo hypotheses were supported, our third hypoth-esis was that Savannah sparrows would choosethe lesser of two evils and preferentially nest inareas of high gull density.

METHODS

Study Site'

Our study site was on Kent Island, an isolated80-ha island in the Bay of Fundy, NewBrunswick, Canada (44°35' N, 66°46' W). Thestudy was centred in two open fields whereSavannah sparrows nest at high densities. TheSouth Field site is a 6-ha portion of an extensiveopen habitat which supports a large breedingcolony of herring gulls (Cannell & Maddox 1983).The North Field is a 1.3-ha field surrounded bywhite spruce, Picea glauca, located several hun-dred metres from the South Field. The fields aredivided by mowed paths into 43 quadrats of0.25-ha (50 x 50 m). Approximately one-thirdof the Kent Island Savannah sparrow popu-lation breeds in the South and North Field studyarea (Dixon 1978; N. T. Wheelwright, personalobservation).

Study Species

On Kent Island, Savannah sparrows (hereaftersparrows) establish territories in coastal vege-tation, marshes, low woody vegetation and openfields. By 1 year of age, both male and femalesparrows are reproductively mature. Their nests,which are built directly on the ground by thefemale, are exceedingly well concealed, typicallycovered by vegetation and accessible only by anarrow grass-lined tunnel (Wheelwright & Rising1993). By the time the sparrows return from theirsouthern wintering grounds in early May, gullsare already present and preparing to nest, so thedensity of gulls within a sparrow's prospectiveterritory can be easily assessed early in the season.

Female sparrows appear to be relatively free tochoose nest sites in different habitats. Breedingdispersal by males and females is negligible, how-ever; once a female nests in a particular location,she generally locates future nests near her originalsite regardless of the density of sparrows or gulls(N. T. Wheelwright & R. A. Mauck, unpublished

data).Each summer from 1987 to 1994, all Savannah

sparrows in the South and North Fields werebanded with U.S. Fish and Wildlife Service alu-minium bands and a unique, randomly deter-mined combination of three plastic colour bands.During daily censuses, the sparrows' behaviourpatterns (singing, foraging, mate guarding, feed-ing nestlings or fledglings) were recorded on fieldmaps (scale I: 1200). All but a few sparrow nestswithin the study area were located each year(Wheelwright & Schultz 1994). Nest sites weredesignated with 6.3 x 8.8-cm vinyl flags on wirestakes placed 3 m south of the nest itself; weinconspicuously bent several blades of dry grassor goldenrod (Solidago spp.) stems directly overthe nest to help us relocate it. As a precautionagainst predators learning to associate flags withnests, we placed dummy flags throughout thestudy area. We found no difference in predationrate between marked and unmarked nests (chi-square test: P>0.50; N= 16 nests). Nests werechecked every other day until the first nestlingshatched and then not again until the nestlingswere 7 days old (2 days before fledging), where-upon the nestlings were measured and banded.With no mammalian predators on Kent Island,the frequency of our nest visits probably didnot appreciably increase predation risks (Major1990; N. T. Wheelwright, unpublished data).We also recorded the fate of each nest (fledgedsuccessfully, abandoned or preyed upon duringincubation or the nestling period).

Herring gulls and American crows are by farthe most important predators of Savannah spar-row eggs, nestlings and fledglings on Kent Island.They are two of the most abundant species ofpredators on the island, with population sizes ofabout 9000 (Cannell & Maddox 1983) and 50 (N.T. Wheelwright, personal observation); respect-ively, and both species readily eat sparrows at allstages of development (based on direct obser-vations and the discovery of fledgling and, in-frequently, adult sparrow bands in herringgull pellets; see also Dixon 1978). Gt;eater~

Animal Behaviour, 53, 1200

black-backed gulls, L. marinus, and commonravens, C. corax, nest on Kent Island and mayopportunistically prey on the eggs and nestlings ofsmall birds, but they are much less common thantheir congenors (about 40 and 8 individuals,respectively). Other potential avian predators ofSavannah sparrows occur on Kent Island onlyduring migration. No reptiles have been found onthe island, and none of the island's few mammalspecies (snowshoe hares, Lepus american us, musk-rats, Ondatra zibethica, and bats, Myotis spp.)preys on nests.

Observations of the foraging behaviour of her-ring gQlls (hereafter gulls) and crows suggestedthat crows pose a greater threat to sparrows eventhough gulls are much more common withinthe study area. Gulls spend much of the day neartheir nests, guarding or incubating their own eggs(Pierotti & Annett 1991) or interacting with othergulls rather than hunting for the eggs of otherspecies (Pierotti 1983). Most of their foraging isdone in the inter-tidal zone or at sea, althoughthey sometimes hunt in the fields for insects. Gullsapparently only prey on the cryptic Savannahsparrow nests when they happen upon them intheir search for insects or incidentally flush theincubating female (cf. Vickery et al. 1992). Spar-row nests within 1 m of active gull nests routinelyfledged young, despite the fact that parentsparrows made more than 200 conspicuous feed-ing trips per day during the nestling period(Wheelwright et al. 1992). Crows, on the otherhand, typically forage methodically in smallgroups which sweep through the fields. Crows arealso proficient at remembering the locations ofindividual nests (Sonerud & Fjeld 1987). Bothgulls and crows avoid walking through densepatches of goldenrod and raspberry; Rubus idaeus,but crows (unlike gulls) habitually fly betweenlowbush blueberry patches, Vaccinium angustifo-lium, a popular nesting habitat of Savannah spar-rows, and land on them in a systematic search fornests (see also Sullivan & Dinsmore 1990). In thecases where we have watched crows hunt for nestsand eat sparrow eggs or nestlings, they appearedto cue in on sparrow alarm calls, narrowing theirsearch as the intensity of alarms increased (N. T.Wheelwright & J. Mitchell, unpublished data).Crows will also steal eggs from gull nests (Ewins1991; N. T. Wheelwright, personal observation),which presumably explains why gulls attackcrows near their nests and why we rarely noticed

crows foraging in the immediate vicinity of nest-

ing gulls.Between 1987 and 1994, Savannah sparrows

nested at a mean :i: SD density of 6.2 :i: 6.1 females/ha in the study site. Densities varied betweenquadrats, however, with 0-32 females ne'sting perquadrat (ANOV A; P

Wheelwright et al.: Sparrow nest-site selection

201

the incubation shift was completed and the femalehad left the nest, the model was removed and thesame observations were made without the model

present.A similar procedure was followed during the

nestling period. To avoid habituation to themodels, a female that was exposed to a crowmodel during incubation was presented with agull model during the nestling period, and viceversa. At each nest, we recorded the nUmber oftimes that the male and female sparrow arrivedcarrying food, the length of time between arrivaland actual entrance of the nest and the method ofapproach, as described above. After 30 min, themodel was removed and the same observationswere made without the model present.

We performed ANOV As to determine the effectof model type (crow versus gull), time of season{early versus late clutches), the stage of reproduc-tion (incubation versus nestling period), andlocation of nest-site (quadrats where gull densitywas zero-low versus intermediate-high) on thebehavioural variables described above.

Finally, we exposed gulls to crow models placed5 m from their nests and recorded their behaviourfrom at least 30 m away. The time that it tookgulls to respond to the models as well as thenature of their response were recorded during theincubation and nestling periods. Except wherenoted otherwise, descriptive data are presented asmeans :1: SD. Statistical analyses were performedusing Statview (Abacus Concepts 1992).

placed Savannah sparrow nests, each contain-ing four eggs, near gull nests to monitor eggdisappearance rates in early June 1991. The nestsand eggs (of tree swallows) had been saved froma separate study on age-related reproduction(Wheelwright & Schultz 1994); tree swallow eggsare similar in size to those of Savannah sparrows,but they are pure white rather than dull brownwith speckles. The sparrow nests were hidden innatural vegetation in matched pairs at distances of1 m and 5 m from each of 10 gull nests (N=20artificial nests).

To test sparrows' reactions to different preda-tors, we presented 37 breeding female Savannahsparrows with life-size standing plastic models ofherring gulls and American crows (Carry-LiteInc., Milwaukee, Wisconsin, U.S.A.). As a con-trol, we initially used a model flamingo, which wassiri1ilar in size to both the gull and crow models.The sparrows paid little attention to it, however,so we used no model for the rest of the controls.Experiments were conducted halfway throughboth the incubation and nestling periods in allweather except heavy rain, as described below.Parental age, clutch size and brood size did notvary between treatments (ANOY A; P>0.30,N=37 females).

In experiments conducted during the incubationperiod, one of the model predators (crow or gull)was randomly selected and concealed in a baguntil we placed it in a realistic position on theground 5 m from a Savannah sparrow's nest. Wemade all behavioural observations using a 15xspotting scope and lOx binoculars while hiddenin: a blind or behind tall vegetation at least25 m from the nest and 30 m frbm the model(Wheelwright et al. 1992). After the femalesparrow appeared near her nest, we recorded thenumber of alarm calls, the length of time betweenher arrival and actual entrance to the nest, hermethod of approach (flying directly to the nest,flying to a perch within 1 m of the nest andwalking the rest of the way, or walking from adistance of > 1 m), the closest distance that thefemale approached the model, and the length ofher incubation shift after she entered the nest. Thetime that a female actually entered her nest wasestimated by extrapolating from a female's speedover the ground, noting the time when vegetationstopped moVing, and listening for the cessation ofsoft alarm notes. If the female did not enter thenest within 30 min, the model was removed. After~

RESULTS

Nest-site Selection by Savannah Sparrows

Savannah sparrows nested at higher densities inquadrats where gull density was relatively low.The density of first-clutch nests (which corre-sponds to the number of individual breedingfemales) in quadrats where gulls were absent wasnearly twice that of quadrats where gull densitywas high, despite similarities in vegetation struc-ture (zero gull density: 7.9:1: 7.7 sparrow nests/ha;low gull density: 6.1:1: 5.9; intermediate gull den-sity: 5.3 :I: 4.4; high gull density: 4.3:1: 4.5; one-way ANOVA; P=O.OI7). Sparrow nest densityalso varied between years, but the negative effectof gull density on the density of sparrow nestsremained significant when controlling for year(two-way ANOV A; year: P

202 Animal Behaviour, 53, 1

Table I. Relationship between gull density near a Sav-annah sparrow's natal nest and gull density near its firstadult nest. For neither females nor males was there asignificant relationship, although male sparrows showeda slightly greater tendency than females to nest in areaswith gull densities similar to their natal nest site (chi-square test: females: P=0.70; males: P=0.09)

Gull densitynear adultnest

Gull density near natal nest

Zero-low InterDlediate-high

2214

147

265

1510

FemalesZero-lowIntermediate-highMalesZero-lowIntermediate-high

rounding their nests was about 4.5 cm taller, onaverage, than that of sparrows breeding in areasof low gull density. Sparrow nests in areas ofintermediate-high gull density were more likely tobe open rather than covered (like greater vege-tation height, a consequence of placing their nestsamong blueberry and goldenrod).

Sparrows nesting among gulls also seemedunusually wary in approaching their nests. Theytended to give more alarm calls and were oftenindirect in approaching their nest, flying to within1 m of the nest and then walking the rest of theway.

Sparrows nesting near gulls spent more timeon nearby perches before entering the nest to feedtheir young. Near gulls, under natural conditions,sparrows perched for 7.1 :I:: 4.2 min before enter-ing the nests, compared with 5.1:1:: 4.2 min forsparrows far from gulls (ANOV A; P=0.07; N= 59l-h observation periods). Feeding rate did notvary as a function of gull density (22.1 :I:: 6.0 trips!h near gulls versus 21.3:1:: 2.8 trips/h away fromgulls; ANOV A; P=0.76). One consequence of thecautious behaviour of sparrows nesting near gullswas that their cryptic nests were more difficult forus to find. In areas where gulls were absent, wefound nests an average of 6.3:1:: 5.0 days (N= 163)before hatching; where gull densities were low, theinterval between discovery and hatching was simi-lar (6.4:1::5.3 days, N=lll). Where gulls werecommon, on the other hand, we sometimes didnot discover nests until shortly before hatching(intermediate gull densities: 4.5:1:: 5.7 days,N= 110; high gull densities: 3.0:1:: 5.7 days, N=97;ANOV A; P0.42for both sexes). Likewise, there was no relation-ship between gull density and various measures ofbody size (wing length, tarsus length, body mass,bill length, bill depth; one.way ANOVA; P>O.lOfor all comparisons) of nesting sparrows. Therewas

also no indication that sparrows that nestedamong gulls were late arrivals from spring migra-tion faced with fewer choices of nest sites. Themean date of hatching of the first clutch in quad-rats where gull densities were zero, low or highdiffered by only 0.3 days (N=72, 40 and 35 nests,respectively). The mean date of hatching in quad-rats with intermediate gull densities (N=40 nests)was

less than 2 days later than in other quadrats,and overall there was no significant relationshipbetween gull density and date of hatching (one-way

ANOV A; P=0.5l; two-way ANOV A; year:P=0.09; gull density: P=0.46; year x gull density:

P=0.99).To determine whether the choice of nest sites by

sparrows was influenced by early experience or apossible genetic predisposition, we examined therelationship between gull densities near an indi-vidual's

natal nest and its first adult nest. Bothfemale and male sparrows appeared to choosenest sites independently of their natal nest site,although our data suggested that males maybe more likely than females to select breedingsites with gull densities similar to their natal site

(Table I).The behaviour patterns of sparrows nesting in

areas where gulls were common were distinctive inseveral respects. Compared with sparrows breed-ing in areas where gulls were absent or scarce,sparrows nesting near gulls disproportionatelyselected nest sites in dense patches of lowbushblueberry and goldenrod. The vegetation sur-~

Wheelwright et al... Sparrow nest-site selection 203

were hard to find was that sparrows nestingamong gulls took advantage of alarm calls bygulls and frequently flushed from their nests assoon as the gulls warned of our approach (e.g.Burger 1984).

Reproductive Consequences of Nest-site Selection

To control for site effects as well as seasonalchanges in vegetation height and the foragingbehaviour of gulls and crows, we focused on firstclutch nests in the South Field study site (wefound qualitatively similar results when werepeated the analysis with the entire sample ofnests). Predation rates did not differ significantlybetween years (ANaYA; P>O.IO), so wecombined data from all 8 years of the study.

Of 163 nests located in quadrats where gulldensities were intermediate-high, 70.7% fledged atleast one young. In contrast, only 55.7% of nestsfledged young in quadrats where gull densitieswere zero-low (N= 103; chi-square test; P=0.02).Merely tallying up the proportion of nests thatfledged young can give a biased picture of actualpredation risks, however, unless all nests aremonitored for the same length of time, which wasnot the case in our sample because of the difficultyin finding nests near gulls (see above; Mayfield1975; Hensler & Nichols 1981). We correctedfor different amounts of exposure to predatorsby calculating daily survival rates during theincubation and nestling periods.

There was no difference in the probability ofpredation during the incubation period in areaswhere gulls were absent or scarce coIppared withareas where gulls were relatively common. Dailysurvival rates where gull densities were zero-lowaveraged 0.945:f: 0.008 (N=167 nests). In quad-rats where gull densities were intermediate-high,daily survival rates averaged 0.934 :f:, 0.014 (N=99nests; P>O.IO using Hensler & Nichols' (1981)Mayfield test statistic). The overall probability ofsurvival over the entire l2-day incubation periodwhere gull densities were zero-low was 0.504(0.94512) and 0.443 where gull densities were

intermediate-high.During the nestling period, the intensity of

predation declined markedly compared with theincubation period. Nests away from gulls hadsignificantly lower daily survival rates than thosewhere gulls were common (zero-low gull densities:0.972:f: 0.005, N= 120 nests; intermediate-high

gull densities: 0.989::i: 0.004, N=77 nests, P

204 Animal Behaviour, 53, 1

removed from any of the artificial sparrow nestsafter 12 days, regardless of whether the exper-imental nests were located 1 m or 5 m from anactive gull nest. When eggs were presented in anobvious way to gulls, or if the gulls happenedupon a nest, the eggs were readily consumed(N. T. Wheelwright, personal observation).

Behavioural Responses to Potential Predators

Savannah sparrows reacted to the presence oflive and model predators near their nests byperching at a distance of 1-5 m from the predatorand uttering alarm calls, a series of high-pitchednotes at intervals of 1 s or less (Wheelwright &Rising 1993). As potential predators approachedthe nest, the interval between notes becameshorter and the calls became louder and higherpitched. When a predator was near their nests,birds erected their crest feathers and often flewaround the nest area, occasionally dropping to theground and walking about before perching againto observe the potential predator. Males some-times sang softly until the female entered the nestto incubate.

Control Gull CrowModel

Figure 1. Mean:i: SD amount of time (min) that femaleSavannah sparrows spent before entering their nestsafter detecting a model predator placed 5 m from thenest. ~parrows delayed signifi~tly longer in the pres-ence of a crow model than a herring gqil model orcontrol (ANOY A; PO.IO).N trials given above error bars.

Responses to Model Predators

The amount of time it took female Savannahsparrows to enter the nest in the presence of amodel predator did not vary as a function of thestage of the reproductive cycle (incubation versusnestling period; ANOV A; P=0.93, N= 146 trials;Fig. 1). Therefore, in the following analyses, wecombined trials conducted during the incubationand nestling periods. The results presented belowwere similar even when incubation and nestlingperiods were considered separately.

The lack of a difference between the incubationand nestling period indicated that the hesitancesparrows showed",was related more to an avoid-ance of revealing their nest's location (or perhapsbeing caught themselves on the nest) than todefence of their offspring per se (see below). Thestage of young had less influence on the responsesof adult sparrows to models than the type ofpredator. Our experiments were not designed totest parental responses to changes in investment inor reproductive value of offspring (e.g. Gottfried1979; Weatherhead 1979; Patterson et al. 1980;Montgomerie & Weatherhead 1988; Redondo1989).

There was no effect of the type of predator onthe number of alarm calls given by sparrows(Mann-Whitney U-test; P=0.47), the length of afemale's incubation shift once she entered the nest(in the presence of the gull model: 21.2:!: 10.4 min;crow model: 19.3:!: 10.8 min; P=0.50), or thecloseness of her approach to the model (minimumproximity to the gull model: 3.3:!: 1.4 m; to thecrow model: 2.8:!: 1.6 m; P=0.32). Femalesdelayed longer before entering their nests if a gullmodel was present than if no predator was present(ANOV A; P

Wheelwright et al.: Sparrow nest-site selection 205

nesting where gull densities were zero-low(ANOV A; P=O.O9, N= 146).

Figure 2. Percentage of nest entrances that were direct(as opposed to birds flying to a nearby perch andwalking inconspicuously the rest of the way) in femaleSavannah sparrows when model predators were placed5 m from the nest. Sparrows were less likely to entertheir nests directly in the presence of crow models thanin the presence of gull models (chi-square test; P

206

Animal Behaviour, 53, 1

sometimes nested successfully within 1 m of gullnests. We know of no reported cases of songbirdsnesting conspicuously and successfully so close tocrow nests, and egg predation has been shown tobe higher near crow nests than away from them(Slagsvold 1980; Nilsson et al. 1985). Our obser-vations and experiments demonstrated that gullsaggressively chased crows and other sparrowpredators from the vicinity of their nests. Spar-rows nesting near gulls were able to compensatefor the elevated risk of gull predation on their eggsor nestlings by bwlding their nests in micro-

,

habitats that gulls rarely frequented, such asgoldenrod and blueberry patches. 'In the presenceof a predator, they were also more cautious aboutentering their nests and more likely to approachthe nest indirectly, presumably to reduce the riskof revealing the location of their nests. This studydemonstrated that nesting near gulls may reducepredation on a sparrow's eggs and nestlings, butfor adult sparrows themselves, there may bebenefits in using gulls as shields or early warningsystems against hawks, owls or other predators.

Given the net benefits of nesting near gulls, onemight predict that sparrows would prefer suchnest sites, and that, as a result, the density ofnesting sparrows would be higher near gulls. Wefound that sparrows tended to avoid nesting neargulls, however. We can eliminate some expla-nations for this apparent paradox. First, gull den-sity near a Savannah sparrow's natal nest and itsadult nest were not correlated; i.e. the tendency tonest near gulls was apparently not a heritable traitor one influenced by a bird's experience as a nes-tling. Second, sparrows nesting near gulls were notnecessarily inexperienced or subordinate to birdsnesting away from gulls. They were indistingwsh-able in a variety of traits related to social status(body size, age, date of nesting). Another indi-cation that nesting among gulls was not simply acharacteristic of younger or subordinate birds wasthe fact that, once a sparrow selected a territory orgeneral breeding site, it showed strong breedingphilopatry, moving a median distance of less than30 m between years regardless of gull density;males and females were equally philopatric (N. T.Wheelwright & R. A. Mauck, unpublished data).

period than sparrows nesting away from gulls, norwas their own survival or the condition of theirfledglings lower. In fact, during the nestling periodnesting near gulls was associated with significantlylower predation rates, which resulted in greateroverall reproductive success.

Our data suggest that lower rates of predationon sparrow nests near gulls was mainly due to theprotection that gulls provided from crows. Wepropose

that sparrows had higher reproductivesuccess near gulls because the gulls, in defendingtheir own nests from predatory crows, effectivelyshielded the sparrows from more dangerouspredators. The gulls' alarm calls in response to theapproach of potential predators such as crows(as well as greater black-backed gulls, ravensand humans) also alerted incubating sparrows,enabling them to flush early from their nests andmaking the nests more difficult to find. We cannotrule out the possibility that the greater nest den-sity of sparrows in habitats with few gulls madesuch areas more profitable places for crows tosearch. Such an effect would have resulted inhigher nest-loss rates in such areas independent ofdirect protectiofi by gulls (Martin 1988, but seeAndren 1991).

Experiments with models demonstrated thatsparrows recognized gulls as potential predatorsbut considered crows a far greater threat, a dis-tinction that was consistent with our observationsthat crows were more systematic in their nesthunting (N. T. Wheelwright & J. Mitchell, unpub-lished data). Certain aspects of the sparrows'precautions in the presence of different modelpredators (the similarity in alarm-calling ratesbefore entering the nest and in the length ofincubation shifts after entering the nest, the closeproximity of their approaches to the models)indicated that the sparrows did not perceive of themodels as posing a risk to themselves. This is inaccord with the fact that we have never seen gulls,crows or ravens prey on adult birds from theground. None the less, the sparrows reacted asthough the model presented a risk to their eggs ornestlings, and behaved differently in the presenceof different model predators.

Patterns of nest predation implicated crows asmore important predators than gulls (e.g. heavylosses in blueberry patches, which gulls avoided,and the failure of gulls to discover artificial nests).Another indication of the relative ineffectivenessof gulls as predators is that Savannah sparrows

Constraints on the Evolution of Nest Associations

The existence of protective nest associationsbetween birds and their predators has been

Wheelwright et al.: Sparrow nest-site selection 207

documented in a few cases (Burger 1984, Ueta1994). What remains to be established is whetherthe behaviour patterns leading to such associ-ations represent true adaptations. Although selec-tively advantageous under certain circumstances,such associations probably only rarely evolve. Inour study, a predisposition to nest near gullspresumably has not evolved in Savannah spar-rows because the heritable basis for such a traitappears to be low and the strength of directionalselection favouring it relatively weak. Avoidinggulls and other large predators is almost certainlystrongly an adaptive behaviour throughout mostof the range of Savannah sparrows.

Gene flow from populations where predatoravoidance is always favoured could easily overrideweak selection for nesting among predators at aparticular site such as Kent Island. At most timesin the lives of most birds, predators are a serioushazard, although they might provide a protectionumbrella near the nest site (Drycz et al. 1981). Thedecision to nest near a predator requires overcom-ing innate and adaptive fears. With life expectan-cies of less than 2 years (Wheelwright & Rising1993), Savannah sparrows and other short-livedbirds may have insufficient time to learn that theenemy of one's enemy is one's friend.

ACKNOWLEDGMENTS

For field assistance we thank R. Anderson,C. Detwei11er, J. Devine, C. Freeman, D.

Harrington, P. Hodum, R. Ingalls, B. McKnight,M. Murray, R. Rynning, C. Schultz, J. Sevigny,

G. Trussell and G. Wheelwright. C. Detweillercollected data on feeding rates in 1988, J. Mitchellobserved crow behaviour in 1995, C. Huntington,S. Johansen, R. Pierotti and an anonymousreferee made helpful comments on earlier versionsof the manuscript, and R. Mauck's computerprogramming was indispensable for data analysis.N.T.W.'s

research has been supported by theNational Science Foundation and a Fulbrightgrant at the University of Botswana. This repre-sents Contribution No. 122 from the BowdoinScientific Station.

REFERENCES

Alberico, J. A. R., Reed, M. J. & Oring, L. W. 1991.Nesting near a common tern colony increases anddecreases spotted sandpiper nest predation. Auk, 108,904-910.

Andren, H. 1991. Predation: an overrated factor forover-dispersion of birds' nests? Anim. Behav., 41,1063-1070.

Bourget, A. A. 1973. Relations of eiders and gullsnesting in mixed colonies in Penobscot Bay, Maine.Auk, 90, 809-820.

Blus, J. J. & Keahey, J. A. 1978. Variation in repro-ductivity with age in the brown pelican. Auk, 95,128-134.

Burger, J. 1984. Grebes nesting in gull colonies: protec-tive associations and early warning. Am. Nat., 123,327-337.

Cannell, P. & Maddox, D. 1983. Population changein three species of seabirds at Kent Island, NewBrunswick. J. Field Ornithol., 54, 29-35.

Cullen, J. M. 1960. Some adaptations in the nestingbehaviour of terns. Proc. Int. Ornithol. Congr., 12,153-157.

Curio, E. 1975. The functional organization of anti-predator behaviour in the pied flycatcher: a study ofavian visual perception. Anim. Behav., 23, 1-115.

Curio, E. 1988. Cultural transmission of enemy recogni-tion by birds. In: Social Learning (Ed. by T. Zentall &B. G. Galef, Jr), pp. 75-97. Hillsdale, New Jersey:Lawrence Erlbaum.

Dixon, C. L. 1978. Breeding biology of the Savannahsparrow on Kent Island. Auk, 95, 235-246.

Drycz, A., Witkowski, J. & Okulewicz, J.. 1981. Nestingof 'timid' waders in the vicinity of 'bold' ones as anantipredator adaptation. Ibis, 123, 42~5.

Dwernychuk, L. W. & Boag, D. A. 1972. Ducks nestingin association with gulls: an ecological trade-off. Can.J. Zool., 50, 559-563.

Erwin, R. M., Galli, J. & Burger, J. 1981. Colony sitedynamics and habitat use in Atlantic coast seabirds.Auk, 98,550-561.

Ewins, P. J. 1991. Egg predation by corvids in gullcolonies on Lake Huron. Colon. Waterbirds, 14, 186-189.

Gerell, R. 1985. Habitat selection and nest predation ina common eider Somateria mollissima population insouthern Sweden. Ornis Scand, 16, 129-139.

Gotmark, F. & Ahlund, M. 1986. Nest predation andnest site selection among eiders Somateria mollissima:the influence of gulls. Ibis, 130, 11-123.

Gottfried, B. M. 1979. Anti-predator aggression in birdsnesting in old field habitats: an experimental analysis.Condor, 81, 251-257.

Hensler, G. L. & Nichols, J. D. 1981. The Mayfieldmethod of estimating nest success: a model, estimatorsand simulation results. Wilson Bull., 93, 42-53.

Joyce, F. 1993. Nesting success of rufous-naped wrens(Campylorhynchus rufinucha) is greater near waspnests. Behav. Ecol. Sociobiol., 32, 71-77.

Knight, R. L. & Temple, S. A. 1986. Why does intensityof avian nest defence increase during the nestingcycle? Auk, 103, 318-327.

Abacus

Concepts. 1992. Statview Berkeley, California:Abacus Concepts, Inc.

Animal Behaviour, 53, 1208

Sonerud, G. A. & Fjeld, P. E. 1987. Long-term memoryin egg predators: an experiment with a hooded crow.Omis Scand, 18, 323-324.

Slagsvold, T. 1980. Egg predation in woodland in re-lation to the presence and density of breeding field-fares Turdus pilaris. Ornis Scand., 11, 92-98.

Sullivan, B. D. & Dinsmore, J. J. 1990. Factors affectingegg predation by American crows. J. Wildl. Mgmt, 54,433-437.

Ueta, M. 1994. Azure-winged magpies, Cyanopica cya-nea, 'parasitize' nest defence provided by Japanese lesssparrowhawks, Accipiter gularis. Anim. Behav., 48,871-874.

Vickery, P. D., Hunter, M. L., Jr & Wells, J. V. 1992.Evidence of incidental nest predation and its effectson nests of threatened grassland birds. Oikos, 63,281-288.

Walters, J. R. 1990. Anti-predator behavior of lapwings:field evidence of discriminative abilities. Wilson Bull.,102,49-70.

Weatherhead, P. J. 1979. Do Savannah sparrows com-mit the Concorde Fallacy? Behav. Ecol. Sociobiol., 5,373-381.

Wheelwright, N. T. & Rising, J. D. 1993. Savannahsparrow (Passerculus sandwichensis). In: The Birds ofNorth America, No. 45 (Ed. by A. Poole & F. Gill).Philadelphia: The Academy of Natural Sciences;Washington, D.C.: The American Ornithologists'Union.

Wheelwright, N. T. & Schultz, C. 1994. Age and repro-duction in Savannah sparrows and tree swallows.J. Anim. Ecol., 63, 686--702

Wheelwright, N. T., Schultz, C. B. & Hodum, P. 1992.Polygyny and male parental care in Savannah spar-rows: effects on female fitness. Behav. Ecol. Sociobiol.,31, 279-289.

Wheelwright, N. T. , Trussell, G. C., Devine, J. P. &Anderson, R. 1994. Sexual dimorphism and sexratio in juvenile Savannah sparrows (Passerculussandwichensis). J. Field. Ornithol., 65, 520-529.

Winkler, D. W. 1992. Causes and consequences ofvariation in parental defense behavior by tree swal-lows. Condor, 94, 502-520.

Young, B. E., Kaspari, M. & Martin, T. E. 1990. Speciesspecific nest site selection by birds in ant acacia trees.Biotropica, 22, 310-315.

Koskimies, J. 1957. Terns and gulls as features ofhabitat recognition for birds nesting in their colonies.Ornis. Fenn., 34, 1-6.

Kruuk, H. 1964. Predators and anti-predator behaviourof the black-headed gull (Larus ridibundus L.). Behav.Suppl., 11, 1-129.

Lessells, C. M. & Krebs, J. R. 1989. Age and breedingperformance of European bee-eaters. Auk, 106, 375-382.

McCain, J. W. 1975. A vegetational survey of thevascular plants of the Kent Island group. Rhodora, 77,196-209.

Major, R. E. 1990. The effect of human observers on theintensity of nest predation. Ibis, 132, 608-612.

Martin, T, E. 1988. On the advantage of being different:nest predation and the coexistence of bird species.Proc. natn Acad. S(:i. U.S.A., 85, 2196-2199.

Mayfield, H. F. 1975. Suggestions for calculating nestsuccess. Wilson Bull., 87, 456--466.

Montgomerie, R. D. & Weatherhead, P. J. 1988. Risksand rewards of nest defense by parent birds. Q. Rev.Bioi., 63, 167-187.

Nice, M. M. & ter Pe1kwyk, J. 1941. Enemy recognitionby the song sparrow. Auk, 58, 195-214.

Nilsson, S. G., Bjorkman, C., Forslund, P. & Hoglund,J. 1985. Egg predation in forest bird communities onislands and mainland. Oecologia (Berl.), 66, 511-515.

Nuechterlein, G. L. 1981. Information parasitism inmixed colonies of western grebes and Forster's terns.Anim. Behav., 29, 985-989.

Patterson, T. L., Petrinovich, L. & James, D. K. 1980.Reproductive value and appropriateness of responseto predators by white-crowned sparrows. Behav. Ecol.Sociobiol., 7, 227-231.

Pierotti, R. 1983. Gull (Larus argentatus) puffin inter-actions on Great Island, Newfoundland. Bioi. Con-serv., 26, 1-14.

Pierotti, R. & Annett, C. A. 1991. Diet choice in theherring gull: constraints imposed by reproductive andecological factors. Ecology, 72, 319-328.

Redondo, T. 1989. Avian nest defense, theoreticalmodels and evidence. Behaviour, Ill, 161-195.

Reese, K. P. & Kadlec, J. A. 1985. Influence of highdensity and parental age on the habitat selectionand reproduction of black-billed magpies Pica picahudS'onia. Condor, 87, 96-105.

Reid, M. L. & Weatherhead, P. J. 1990. Mate-choicecriteria of Ipswich sparrows: the imp9rtance of vari-ability. Anim. Behav., 40, 538-544.