j. Field Ornithol., 56(3):236-245

NEST MATERIAL PREFERENCES OF GREAT

BY LINCOLN FAIRCHILD, SHEILA A. MAHONEY, AND RALPH W. SCHREIBER Analysesof variouspasserine nests (Kilgore and Knudsen 1977, Collias and Victoria 1978, Jacobset al. 1978, Thomas 1983) and laboratory studies(Sargent 1965) have shownsome nesting material preferences. No suchstudies have been done on marine ,and none has com- pared selectednesting material to material availabilityin the environ- ment. We comparedthe nesting material in nestsof Great Frigatebirds (Fregataminor) on EasternIsland, Midway Atoll, PacificOcean, to that availablein the environment.We experimentallytested whether Great Frigatebirdsprefer certain typesof sticksor merely take sticksin pro- portion to their availability.

METHODS We conductedour study during May and early June 1983, in the frigatebirdcolony located on the windward,south central side of Eastern Island, Midway Atoll which contained76 nestsbuilt in the crownsof Scaevolabushes about 2 m above the ground. Inter-nest distancesin the sameclump were highly variable(approximate range 1 to 10 m). At the beginningof the study,most nesting frigatebirds were incubating(n = 59), but 5 pairs brooded chicks;2 pairs had just formed and had not producedeggs; and 10 maleswere still displaying.By the end of the study,most nests still containedeggs (n = 36), but 22 nestscontained chicks,and no maleswere displaying.Thirteen nestshad been aban- doned. We examined 6 recently abandonednests. Five of the 6 nestswere examinedimmediately after abandonmentprior to any nestdestruction by other frigatebirds.The sixthnest was examined several hours after abandonmentbut appeared intact. We measuredthe length of each stickin each nestto the nearestmillimeter and quantifiedthe sizeand numberof branchesof all sticksby assigninga valueof 1.0 to a branch if the branch was about the same diameter as the main stem of the stick. We assigneda value of .5 to a branch if the branch diameter wasless than the stem diameter. The sum of all values of branches of a stick is the BranchingIndex (BI) of the stick. We log transformedall stick lengthsand branchingvalues. Since the BI of a stick containingno brancheswould be zero and would not allow usto log transformbranch- ing data, we added one to all BI's prior to log transformation.Thus, the BI of a stick without branchesequalled 1.0. If BI -- 2.0, the stick contained either one large branch or two small branches.We used ANOVA to test whether differencesin length or branching existed between nests.

236 Vol.•6, •o. • FrigatebirdNests [237

We surveyedthe ground around the nestingarea for sitesthat con- tained sticks.Open areasthat were smallerin any directionthan about 2 m (the wing spanof frigatebirds)were not consideredbecause frigate- birdsprobably could not easilyland or hover and henceretrieve sticks from them. We sampledsticks on 5 main pathsand 4 fieldsin the nesting area. We deliberatelyselected, as one of the four fields,a sitewith many large sticksthat appearedappropriate for nests. We useda m2 quadrat to sampleground sticksand adopteda sampling regime to determine whether the proximity of the sampleto the edge of the path or field affectedstick distribution. Sticks were sampledevery 10 paces,or approximately6.6 m, along each path: the first sampleon the right side of the path, the secondsample in the path center, and the third sampleon the left sideof the path. This sequenceof sampling wasrepeated along the length of the path and a similarstaggered sam- pling regimewas used for fields:we took samplesalong both sides,along the field'smidline, and half way betweenthe midline and either sideof the field.At eachsample location, we measured and assigned a branching index to all sticksthat were at least 10 cm long and totally or partially withinthe m2 quadrat. Stick lengths and branching were log transformed and analyzedby ANOVA to test whether differencesexisted among quadratswithin a given area. Differencesamong quadrats were found only in the area we deliberatelyselected because of the high densityof large sticks.Excluding this area, quadratswere pooledwithin areasand ANOVA then testedwhether differencesexisted among areas. To test preferencesin sticks,we asked two questions:(1) Is stick positionwith respectto the ground important to frigatebirds?And (2) doesstick type and/or positionwith respectto the groundaffect frigate- stickselection? To answerthe first question,we collectedstraight sticksof variouslengths and dividedthem into 2 groupsequal in number and averagelengths. We painted the tips and middle of the sticksby group for later identification.To control for paint color, the two stick groupsdiffered only in the sequenceof red and blue bandsof paint. We placedall sticksfrom one groupsingly on the topsof Scaevolabushes near the nestingcolony. Sticks from the other groupwere placedsingly on the ground at the edge of a field in front of the samebushes. In the first of two trials, we used32 sticksin each group and left them unat- tendedafter placement.Two dayslater, wesearched the areathoroughly and collectedand countedall remainingsticks. We surveyedfrigatebird nestsand the ground around the nestingarea daily for painted sticks. Our cautiousnest surveys had only a minimal effect on nestingbirds at the possibleexpense of failingto sightpainted sticks that mayhave been incorporatedinto nests.In the secondtrial, we used20 sticksper group and observedall activitynear the bushand groundcontaining our sticks for one hour followinginitial placement. To answerthe secondquestion, in each of 12 trials, we constructed two nestseach with 20 branchedsticks and placedone in a Scaevolabush and the other on the ground in front of the bush. We constructeda 238] L. Fairchitdet at. j. FieldOrnithol. Summer 1985

T^BI•E1. Dimensionsof stickson the groundin pathsand fieldsand in frigatebirdnests. The mean density,length, and branchingindex ___1 standarddeviation are presented.

n Area (quad- n Density Length Branching Site (m2) rats) (sticks) (sticks/m2) (cm) index Path 1 473 15 135 9.0 __+8.5 21.6 __+11.7 1.08 __+1.24 Path 2 466 9 70 7.8 ___5.7 18.6 ___9.8 1.04 ñ 1.19 Path 3 299 15 118 7.9 __+7.8 20.4 ñ 14.2 1.11 __+1.30 Path 4 1090 15 144 9.6 __+6.5 21.1 __+18.4 1.09 __+1.28 Path 5 301 15 148 9.9 __+12.7 18.8 __+11.5 1.07 __+1.25 Field 1 1026 15 234 15.6 ñ 13.8 19.1 ñ 12.5 1.08 __+1.24 Field 2 334 10 95 9.5 _ 11.5 21.3 ñ 10.1 1.06 ñ 1.23 Field 3 452 14 209 14.9 __+12.8 18.2 __+8.3 1.07 __+1.30 Field 4 78 5 271 54.2 __+43.0 21.5 __+14.2 1.07 ñ 1.21 Nest 1 30 41.8 __+13.0 1.98 __+1.72 Nest 2 30 44.9 __+21.7 2.20 --+ 2.06 Nest 3 47 42.5 __+14.5 1.99 ñ 1.88 Nest 4 35 45.9 __+15.3 1.64 __+1.53 Nest 5 48 36.8 __+16.1 1.86 __+1.70 Nest 6 46 39.3 __+15.4 1.90 ñ 1.50 All sites(ground) 4510 113 1424 20.0 --+12.0 1.08 __+1.25 All nests 236 41.4 __+16.1 1.90 __+1.80 third nestof 20 stickscontaining no branchesand placedit on the bush within 2 m of the branchedstick nest. To control for possibledifferences in our selectednest locations, we reversedthe positionof the two nests in the bush in each experiment. All nestswere constructedfrom sticks obtained from abandonedfrigatebird neststo diminish the possibility of using sticksthat were inappropriatefor frigatebirds.We observed thesenests from about 75 m awayfor 1 h and recordedthe numberof landingson eachnest, the cumulativetime spenton each nest,the number of attemptsby birds to remove sticksfrom each nest, and the number of sticksactually removed and carried from each nest. At the end of the hour we disassembled all nests and counted the number of sticksremaining. The above experiment confoundsstick type and apparent nest size. Althoughall nestscontained 20 sticks,branched sticks are more massive than stickswithout branches.Thus, the branched stick nestsappeared more massivethan the straightstick nests. To control for the effect of the apparent nest size in our third experiment, we constructeda small nest with 20 branched sticksand a large nest with 30 stickscontaining no branches.We placedboth nestsin a Scaevolabush within 2 m of each other. About 18 h later, we disassembled the nests and counted the numberof remainingsticks in each.

RESULTS Frigatebirdnests averaged 39.3 ___8.6 sticks(n -- 6 nests).Stick density in all pathswas 8.5 ___8.3 sticks/m2 (n = 69 quadrats),and no significant Vol.56, No. 3 FrigatebirdNests [2 39

45 GROUNDN -- 1424STICKS 40

35 NESTN -' STICKS236

30

25

20

15

10

5

10 15 25 30 35 5 70 75 80 85 86+

STICK LENGTH (cm) FIGURE1. Groundand neststick lengths. The solidbars represent ground sticks, and the hatchedbars represent nest sticks. Sticks are lumpedinto 5-cm sizeclasses for illustrativepurposes only. differenceexisted among the 5 paths(F,64 = .17, P > .05, Table 1). Whilethere was no significantdifference in stickdensity in 3 of the 4 fields(F2.•4 = 1.94,P > .05),the density in thesefields was significantly greaterthan in paths(F,104 = 8.76, P < .005). The averagedensity in fieldswas 13.8 _ 12.8 sticks/m2 (n = 39 quadrats).The averagestick densitywas greater in the fourth field, 54.2 _ 43.0 sticks/m2 (n = 5 quadrats). Averagestick length in nestswas 41.4 _ 16.1 cm (n = 236), whereas the averagelength of groundsticks was only 20.0 _ 12.0cm (n = 1424, Table l) and significantlyshorter than neststicks (F1,1658 = 628.1, P < .001, seeFig. 1). The differencebetween the lengthof groundand nest stickswould be evengreater had we includedin our groundsamples the numeroussticks that were lessthan 10 cm long. Significantlymore branchedsticks were found in neststhan on the ground(X 2 = 586.1, df = 1, P < .001). Only 9.9% of all groundsticks (n = 1410)were branched, whereas 77.5% of neststicks (n = 236) were branched.The averagebranching index of groundsticks was 1.08 _ 1.25 (n = 1410).As expected,large ground sticks were more likely to containbranches than shortersticks (r = .152, df = 1417, P < .01). Whilesignificant, the correlation between length and branching is weak. Groundsticks had significantly fewer branches than neststicks (F1,1644 = 1070.9,P < .001, seeFig. 2). The averagebranching index of nest stickswas 1.90 _ 1.80 (n = 236). There is no significantcorrelation betweenlength and branchingof nest sticks(r = .067, df = 234, P > .05).The averageground stick had no branches.The averagenest stick had at least one major branch or 2 minor branches. 240] L. Fairchildet al. J.Field Ornithol. Summer 1985 In our first experiment, we recovered 28 of the 32 stickscontaining no branchesthat had been placed singly on bushes.Some stickshad fallen to the ground, perhapsindicating unsuccessful attempts by birds to remove them from bush tops. One of 4 missingbush stickswas recovereddirectly beneath a frigatebird nest about 40 m from initial stickplacement. We recoveredall ground sticks,usually exactly where initially placed.After we placed20 stickswithout brancheson a bush, within an hour 2 frigatebirdsflew down, hovered over the bush, and each picked up a stickand flew off with it. No stickswere taken from the ground. During 12 replicatesof the artificial nest experiment, frigatebirds respondedduring 8. In these8 experiments,frigatebird responses began 21.5 ___17.9 min after the nestswere placedand continuedfor 24.1 ___ 18.2 min. Often by the end of the observationperiod one of the nests had been destroyedand stickswere scatteredabout the bush and sur- rounding ground. Frigatebirdsrestricted their activityprimarily to bushnests (Table 2). Only one bird landedon the groundnest, and only one bird attempted to removesticks from the groundnest. On bushes,the total time frigate- birds spent on the straight stick nest was not different from the total time spent with branched sticks.The total number of landingson branched stick nests,however, was about 60% greater than on those nestsconstructed of stickswithout branches.Hence, activityaround the branched stick nest was greater than the activity associatedwith the unbranched stick nest. The number of sticks taken from branched stick nestswas twice the number taken from nestscontaining stickswithout branches, and the total number of attempts to remove sticks from branchednests was also about twice that from straightstick nests. Con- sequently,the effectivenessor the easeof stickremoval wasnot much different. Given the smallsample sizes, however, the aboveresults were not significantlydifferent as determinedby chi-squaretests. These resultsare somewhatmisleading. About 30% of the time, when a bird flew into the area, one nest was already occupied.Thus the incoming bird had a choiceof either selectingthe unoccupiednest or displacingthe bird from the occupiednest. We observedonly one dis- placement.Generally, the incomingbird selectedthe unoccupiednest or flew off. If one considersonly those attempts when all nestswere unoccupied,birds selectedthe branchedstick nest 3 times more often than they selectedthe nestsof stickswithout branches(Table 2), and the differenceis significant(X 2 = 7, df = 1, .005 • P • .01). These behavioraldata correspondwith an analysisof nestcontents. The percent of attempts(75%) by frigatebirdsat branchedstick nests wasnot significantlydifferent from the percentof branchedsticks (77.5 %) found in their nests(X 2= .12, df = 1, P ) .05), but was significantly differentfrom the percentof branchedsticks (9.9%) found on the ground (X2 = 118.6, df = 1, P (.001). In our experiment usingdifferent-sized nests, we recovered 29 of the Vol.56, •o. • FrigatebirdNests [24 1

GROUND STICKS ,oo ß N ----1410 [• NESTSTICKS N = 236

O so

10

1.0 1.5 2.5 3.0 3.5 4.0 4.5 5.0 5.5

STICK BRANCHING INDEX FlouRy.2. Degree of branchingof groundand nest sticks.The solidbars represent groundsticks and the hatchedbars represent nest sticks. The BranchingIndex (BI) combines the number of branches with branch size. If BI = 1.0, the stick has no branches. If BI = 1.5, the stick has one small branch. If BI = 2.0, the stick has one large branchor two smallbranches. See text for explanation.

30 original sticksfrom the large, unbranchedstick nest, but only 7 of 20 branched sticksthat comprisedthe smaller nest. While branched nestswere destroyedin half of the experimentsin which frigatebirds participated,nests of stickswithout brancheswere never destroyed, probablybecause fewer birds approachedstraight stick nests. Even if activitiesbetween nest typeswere similar,branched stick nests would be more prone to destructionthan straight stick nestsbecause in re- moving a branchedstick from a nest, the branchestend to catch other sticksand disrupt them.

DISCUSSION On Midway,frigatebird nests were constructedalmost exclusively of Verbesinasticks. Large fields of Verbesinasurrounded the frigatebird colony, so potential nesting material was abundant. Scaevolaand Cas- uarinasticks were also abundant but not commonlyfound in frigatebird nests.No frigatebirdnest contained sticks that were obviouslycollected from living plants, although Verbesinatassles from living plants were frequentlyused to line the sticknests. All stickswere woody,bleached, 242] L. Fairchilde! al. J.Field Ornithol. Summer 1985

TABLE2. Responsesof Great Frigatebirdsto 3 artificialnests placed near colony.The numbersin the table are the total numberof observationsof the specifiedactivity in 8 replicatesof the experiment.

Straight- Branched- Branched- stick nest stick nest stick nest Activity in bush in bush on ground Total Number of landings 9 15 1 25 Total time birds on nest (min) 23 20 0.5 43.5 Average time on nest/landing (min) 2.6 1.5 0.5 Number of sticks taken 4 8 0 12 Number of attemptsto remove sticks 14 26 1 41 Removal effectivness (no. taken/no. attempts) 0.29 0.31 0 Number of attemptswhen all nestswere unoccupied 7 21 1 29 Number of destroyednests 0 4 0 4 and obviouslydead. Consequently,the year's current crop of Verbesina wasprobably not usedexcept as nest lining. Since a frigatebird nest is composedof about 40 stickson Midway Atoll and sincethe densityof stickswas about 8 sticks/m2 on pathsand about 14 sticks/m2 in fields,a pair of frigatebirdscould theoretically obtain a sufficientnumber of sticksfrom about 5 m= of path or 2-3 m= of field. In 1983, only 76 frigatebirdnests existed on Midway.Since the islandis over 8000 m=, the amountof nestingmaterial per seis obviously not limiting to this population. However, most ground stickswere not suitablenesting material. They were too short and containedtoo few branches.Less than 18% of ground sticksfall within one standardde- viation of the mean length of nest sticks.The average ground stick branching index was about 1.0, whereasthat of nest stickswas nearly 2.0. Lessthan 10% of ground stickswere branched at all. Becausewe sampledground sticksafter most frigatebirdsselected their nestmaterials, we underestimatedthe averagelength and branch- ing index of ground sticksthat were availableto frigatebirds.If we correctedour ground stick samplesto include the nest sticksthat we would have found on the ground prior to frigatebird selection,the averageground stick length would increaseby lessthan 3%, and the averagebranching index would increaseby lessthan 2%. These minor differences due to when stickswere sampled cannot account for the major differenceswe found betweennest and ground sticks. Our stick experimentssuggest that the location of the sticksis im- portant. No stickswere ever taken from the groundin our experiments. All the removedsticks came from the bushesregardless of whether the stickswere placedin their respectivesites singly or clumpedtogether in nest fashion. Despite frigatebirds' preference for stickson bushes, there are very few accessibledead stickson or in busheson Midway. Vol.56, •o. • FrigatebirdNests [243

However,frigatebirds can and do hover directlyabove ground sticks in largeopen areas and pick them up withoutlanding (Nelson 1975, pers. obs.),and this appearsto be the major way frigatebirdsgather nesting material. In our experiments,frigatebirds selected, when a choicewas possible, sticksfrom the branchedstick nest 3 times as often as they selected stickswithout branches.Although the percent of branchedsticks se- lected in these experimentswas the same as the percent of branched sticksfound in nests,there is another possibleexplanation of our ex- perimentalresults. Because artificial nestsof 20 branchedsticks were more massivethan thoseof 20 straightsticks, the preferencemight have been for the seeminglylarger nestrather than for the branchedsticks themselves.However, if nestsize was important, the larger straightstick nestwith 30 sticksshould have been destroyedin our last experiment. Instead we found that the smaller branched stick nest with 20 sticks was destroyed,indicating a clear preferencefor branchedsticks. Our resultsprovide strong evidencethat frigatebirdspreferentially selectlong, branchedsticks as nestingmaterial despite their low avail- abilityon the ground.Branched sticks might be preferred becausethey are easier to anchor in the Scaevola branches and in the nest itself than are stickswithout branches.Longer rather than shorter sticksmay be necessaryto make a nestlarge enoughto accommodatea frigatebird. Like Marsh Wrens (Cistothorusplatensis; Picman and Picman 1980), Great ( carbo; Urban 1979), and Broad-tailed Hummingbirds(Selasphorus platycerus; Calder 1972), frigatebirdsde- stroyedabandoned or unattendedconspecific nests. High mortality of frigatebird eggsand nestlingshas been attributed, in part, to this be- havior (Nelson 1975). Nest robbingmight occurbecause the greatest concentrationof preferred sticks,particularly late in the nestingseason, is found in other frigatebird nests,not on the ground. That nestsprovide the greatestconcentration of nestingmaterial is particularlyimportant becausefrigatebirds are known to breed some- whatasynchronously (Stonehouse and Stonehouse 1963, Diamond 1973, Nelson 1975). Our own observationsthat 10 maleswere displayingin the colonyof 76 pairsand another2 werejust formingpairs while 5 pairswere alreadybrooding chicks indicates nesting asynchrony in the Midwaycolony. Late nestersmight expeditenest construction by rob- bing other nestsof sticksand thusminimizing the searchtime necessary to find suitablesticks. In mostother ,the late nestersare young birds(Coulson and White 1958, Fisher 1969, Mills 1973, Blusand Keah- ey 1978, Ryder 1980, Masseyand Atwood 1981). In our stickexperi- ments,all participantswere males(with the possibleexception of one youngbird whosesex was not determined),and mostof the participants had at least somewhite breast feathersindicating that they had not obtainedmature adult male plumage. These experimentssuggest that frigate birds are activelyselecting specificmaterial for nest building from what is availablein the envi- 244] L. Fairchildet al. j. FieldOrnithol. Summer 1985 ronment.Our data beginto clarify nestconstruction characteristics for this speciesbut do not clarify factorsregulating the frigatebirdpopu- lation on Midway.Further data are neededon factorssuch as the role of bushesas nestingsubstrate, food supply,and climatologicalfactors.

SUMMARY Great Frigatebirdsbuild stick nestsin busheson remote oceanicis- lands,preferentially selecting nest material from whatis available. Nests consistprimarily of long, branchedsticks despite the low availabilityof suchsticks in the environment.In choiceexperiments on MidwayAtoll, GreatFrigatebirds selected branched rather than straightsticks 75% of the time, and theypreferred to pickup nestmaterial from bushesrather than from the ground.The proportionof branchedsticks in frigatebird nestswas not significantlydifferent from the proportionof branched sticksselected in choiceexperiments.

ACKNOWLEDGMENTS We thank Dr. Robert Shallenbergerof the Fishand Wildlife Service, Honolulu, CMD James Harris, and CMD William Parthum, for facili- tating the stayon Midway. We greatly appreciateRussell and Beverly Shea for providing logisticalsupport and companionshipin the field. We thank Drs. M. McVey, J. F. Downhower,and E. H. Burtt, Jr. for their constructivelycritical commentson earlier drafts of this manu- script.We thank The Ohio StateUniversity and FloridaAtlantic Uni- versityfor providingpartial funding for this work.

LITERATURE CITED

BLUS,L. J., ANDJ.A. KEAUEY.1978. Variation in reproductivitywith age in the Brown . Auk 95:128-134. CALDER,W. A. 1972. Piracyof nestingmaterials from and by the Broad-tailedHum- mingbird.Condor 74:485. COLLIAS,N. E., ANDJ. K. VICTORIA. 1978. Nest and material selectionin the Village Weaverbird (Ploceuscucullatus). Anim. Behav. 26:470-479. COULSON,J. C., ANDE. WUITE. 1958. The effectof age on the breedingbiology of the Kittiwake Rissatridactyla. Ibis 100:40-5 t. DIAMOND,A. W. 1973. Noteson the breedingbiology and behaviorof the . Condor 75:200-209. FISUER,H. I. 1969. Eggsand egg-layingin the LaysanAlbatross, Diomedea immutabilis. Condor 71:102-112. JACOBS,C. H., N. E. COLLIAS,ANDJ. T. FUJIMOTO.1978. Nest colour as a factor in nest selectionby femaleVillage Weaverbirds.Anim. Behav.26:463-469. KILGORE,D. L., ANDK. L. KNUDSEN.1977. Analysisof materialsin Cliff andBarn swallow nests:relationship between mud selectionand nestarchitecture. Wilson Bull. 89:562- 571. MASSEY,B. W., ANDJ. L. ATWOOD.1981. Second-wavenesting of the California Least Tern: age compositionand reproductive success.Auk 98:596-605. MILLS,J. A. 1973. The influenceof age and pair-bondon the breedingbiology of the Red-billedGull Larus novaehollandiaescopulinus. J. Anim. Ecol. 42:147-162. NELSON,J. B. 1975. Breedingbiology of frigatebirds--a comparativereview. Living Bird 1975:113-155. Vol.•6, •o. • FrigatebirdNests [245

PIcM^r•,J., ^•D A. K. PICMAN.1980. Destructionof nestsby Short-billedMarsh Wren. Condor 82:176-179. RYDER,J. P. 1980. The influence of age on the breeding biology of colonial nesting seabirds.Pp. 153-168, in J. Burger, B. L. Olla, and H. E. Wind, eds. Behaviorof Marine , Vol. 4, Marine Birds. Plenum Press, New York. S^RCE•T,T. D. 1965. The role of experiencein the nest buildingof the Zebra Finch. Auk 82:48-61. STO•I4OUS•,B., ^r• S. STO•HOUS•. 1963. The frigatebirdFregata aquila on Ascension Island. Ibis 103b:409-422. THOM^S, B. T. 1983. The Plain-fronted Thornbirds: nest construction material choice, and nest defense behavior. Wilson Bull. 95:106-117. U•B^•, E. K. 1979. Observationson the nestingbiology of the Great Cormorantin Ethiopia. Wilson Bull. 91:461-463. Departmentof Zoology,The OhioState Universit),, 1735 Neil Ave., Columbus, Ohio43210 (LF) Departmentof Biological Sciences, Florida Atlantic University, BocaRaton, Florida 33431 (SAM); LosAngeles County Museum of Natural HistoryMuseum, 900 ExpositionBlvd., Los Angeles, California 90007 (RWS). Received8 Jan. 1985; accepted17 June 1985.