j. Field Ornithol., 62(2):256-259

GOLDFINCH PREFERENCES FOR BIRD FEEDER LOCATION

ERICAH. Du• ANDJEREMY A. T. HUSSELL 30 Davidson Road Aurora, Ontario LdG 2B?, Canada

Abstract.--Testswere conductedof AmericanGoldfinch (Carduelis tristis) preference for tube feederswith sunflowerseeds that were simultaneouslyavailable at high and low positionsin a tree, or at low positionsamong tree branchesand in the open. More birds visitedthe upper feederin the heightexperiment, but there was no differencebetween low feedersin and far from the tree. Further study shoulddetermine whether other habitat featuresinfluenced results, and whether goldfinchpreference for high feedersis sufficiently strongto justify recommendationsto feederowners.

PREFERENCIAS DE CARDUELIS TRISTIS POR LA LOCALIZACI•)N DE COMEDEROS Sinopsis.--Se hicieron pruebas para determinar las preferenciasde Carduelistristis para alimentarseen comederoscon semillasde girasol.Dos comederosfueron colocados simul- tfineamenteuno sobre el otroentre las ramas de la partealta y bajade un firboly otrosdos fueroncolocados a 2.4 m de altura, uno entre las ramasde un •rbol y el otro en un hrea abierta.En el experimentode altura, la mayorlade las avesutilizaron el comederolocalizado en el •rea alta; no hubopreferencia de visitasentre el •rea abiertay la cobijadacon ramas. Otros estudiospodrlan determinar si otras particularidadesdel habitat puedenafectar los resultados,y silas preferenciaspor comederoslocalizados en lugaresaltos son lo suficien- temente fuertescomo para justificar recomendacionesalas personasque proveende alimento en comederos a estas aves.

Despitethe popularity of feedingbirds, few scientificstudies have been conductedon the feedersand placementsof feedersthat bestattract favored species.Anecdotal evidence is behind mostof the recommendationsgiven in guidesto bird feeding.Our studywas designedto test the preference of AmericanGoldfinch (Carduelis tristis) for feedersat different heights in a tree, or for feedersat the sameheight hung in a tree or in the open.

METHODS The studytook place during January-March 1990 in Aurora, Ontario (44ø00'N, 79ø30'W), in a suburbanyard with scatteredmature trees.A flock of 15-30 goldfinchesused the test feedersdaily, while only an occasionalvisit was made by 1-3 individuals of other species(Black- cappedChickadee, Parus atricapillus; House Finch, Carpodacusmexican- us; and, rarely, Downy Woodpecker,Picoides pubescens). Two newly-purchasedtube feederswere filled with small black oil sunflowerseed, and hung on a pulley system3.4 and 6.1 m abovethe ground,one feeder directly below the other.Both were amongthe terminal twigs and small branchesof a maple tree (Acer saccamm).The feeders cameequipped with sleevesof plastic-coatedhardware cloth, which served effectivelyas squirrelguards. The lack of perchesand meshsize excluded certain speciesof birds that also frequentedthe study site (European Starling, Sturnusvulgaris; and House Sparrow, Passerdomesticus).

256 Vol.62, No. 2 BirdFeeder Location [257

We recordedthe numberof goldfinchesclinging to eachfeeder during periodicobservations that were made as opportunityarose, but all ob- servationswere separatedby at least 10 min. Between5 and 14 counts were taken during eachtrial (mean = 7.8). Trials endedevery 2-3 d, at which time the amount of food remaining in each feederwas measured, feederswere refilled, and their positionswitched. Six trialswere conducted (three eachwith a particular feederin a particular location). The experimentaldesign paired test locations in eachtrial, controlling for effectsof changingweather, flock size or behavior that might have affected results had tests of each site been made at different times. Switch- ing feedersallowed us to determinewhether one was preferred,regardless of location. A secondset of experiments,using the samemethods, was donewith the feedershung 2.4 m abovethe ground.One was amongtwigs on the outsideedge of a mapletree andthe otherwas 2.4 m awayover a driveway, in the open.As with the height experiment,six trials were run (with feedersswitched at the end of eachtrial). In four additional trials, amount of food was measured,but no bird countswere conducted. Number of birds and total of food used varied widely from trial to trial. To avoid biassingresults towards trials with large numbers,data were convertedto proportions.For eachtrial, the amountof foodremoved from eachfeeder was expressedas the proportionof the foodremoved from both feederscombined (arcsin transformed, Snedecor and Cochrane 1967). Similarly,the averagenumber of goldfinchesat eachfeeder during all countsin a trial wasexpressed as the proportionof the averagenumber seen at both feeders combined. Analysisof variancewas usedto test whether birds preferredone of the two feedersand whether location made a difference (Norusis 1986). If differentspillage rates drew morebirds from onefeeder to the ground, suchvariation would be ascribedby ANOVA to feeder identity rather than to location.

RESULTS Although 15-50 goldfinchesfrequented the feedingarea, the number of birds on a feeder at once usually did not exceedfour. Other birds perchedon twigs or feedersuspension wires near eachfeeder, normally dividedin the sameproportions as the birds actuallyon the feeders. In the heightexperiment, more goldfinches were observed on the higher feeder,but the amountof foodremoved from the upper feederwas larger only when feeder "A" was on top (Table 1). Subsequentexamination showedthat Feeder A had slightly larger holes,probably making food easierto extractand morelikely to spill on the ground.We usedANOVA to controlfor the effectsof feeder identity (A vs. B) while testingthe significanceof feederlocation (high vs.low). Number of birdsand amount of foodremoved were significantlyhigher for the upper feeder,even after controllingfor the effectof which feederwas uppermost(Table 2). In the side-by-sideexperiment, there was an apparenttendency towards 258] E. H. Dunnand ]. A. T. Hussell J.Field Ornithol. Spring 1991 TABLE 1. Percent of birds on, or food removed from, each feeder, accordingto feeder identity and location(average of resultsfor each trial). Percentsare untransformed.

% of birds seen % of food removed Feeder A high Feeder B high Feeder A high Feeder B high Feeder location or in tree or in tree or in tree or in tree

Height experiment High 84.1 67.6 85.6 47.0 Low 15.9 32.4 14.4 53.0 Side-by-sideexperiment In tree 66.6 49.4 74.2 37.0 In open 33.4 50.6 25.8 63.0 preferencefor the feeder in the tree only when feederA was positioned there (Table 1). Significancetests showed that there was no preference for one location over the other (Table 2). There was a much greater proportionof unexplainedvariation in the side-by-sideexperiment than in the height experiment (Table 2).

DISCUSSION Goldfinchesat the study site fed on the groundbelow the feeders,as well as at all test sites.Birds on the groundfrequently flew up to the tree in minor alarms, however, while those at the highest feeder sometimes continued to feed even as the feeder was being lowered for refilling. Moreover, arriving flocksflew first to the top of the tree beforedropping to feeders,and the high feederwas normally visitedbefore flock members spread to other sites. We suspectthe height preferencedocumented in this study was largely a result of self-protectivebehavior. We were surprisedto find no preferencefor the tree site over the open locationduring the side-by-sideexperiment, as the tree site shouldhave offered more protectionand perching sitesfor goldfinches.The impetus for the experimentwas an observationthat few goldfinchesvisited a feeder in anotherpart of the yard, about 10 rn from any tree, and perhapsour experimental feeder was not far enough from the tree to demonstratea real preference. In conductingour study, we became aware of several problems in controlling such experiments.First, we demonstratedthe importance of controllingfor possibledifferences in attractivenessof individual feeders, regardlessof location. Golden-crownedSparrow (Zonotriciaatricapilla) and Fox Sparrow(Passerella iliaca) are known to developpreferences for individual siteson a singlefeeder (Pearson 1979), so in certain typesof experiments(e.g., foodchoice), these preferences must alsobe controlled. We were lucky in havingessentially one speciesvisiting the feeders.If many specieswere present,amount of foodremoved from feederswould be meaninglessin assessingpreferences of individual species,and site preferencecould be influencedby the presenceof other species.Next, it Vol.62, No. 2 BirdFeeder Location [259

TABLE2. Resultsfrom ANOVAs of percentof birdson feederand percentof foodremoved: proportionof variation accountedfor by feederidentity and location,and significance.

Height experiment Side-by-sideexperiment Source of variation % birds % food removed % birds % food removed

Feederidentity 0.09* 0.56*** 0.25 0.48** Feeder location 0.79*** 0.42*** 0.22 0.05 Interactions 0 0 0 0 Main effects 0.89*** 0.98*** 0.47 0.53**

* P < 0.05. ** P < 0.01. *** P < 0.001. is difficultto assessthe effectsof surroundinghabitat. Our resultsmight havebeen affected by the presenceof houses,for example,or by a lack of conifers. Despite these problems,further studiesshould be conducted,both on goldfinchesand on otherspecies. Our experimentscould be repeatedeasily •n a variety of circumstancesto determinewhether goldfinchpreference for high feedersoverrides local habitat conditions. Further studiesmight recordthe number of birds feedingon the ground,to determinewhether the groundis asattractive as aerial feeders.Lastly, it shouldbe determined whetherpreferences for feederlocation are strongenough to justify rec- ommendationsfor feederowners. Possibly the samenumber of birdscould be attracted regardlessof feeder location if no choiceof sitesis offered, althoughthis is difficult to test rigorously.

ACKNOWLEDGMENTS

Thanksto BruceBacon and David Hussellfor their helpfulcomments on our manuscript.

LITERATURE CITED

Noe,usIs, M.J. 1986. SPSS/PC+ for the IBM PC/XT/AT. SPSSInc., Chicago,Illinois. PE^RSON,O. P. 1979. Spacingand orientationamong feeding Golden-crowned Sparrows. Condor 81:278-285. SNEDECOR,G. W., AND W. G. COCHRAN. 1967. Statistical methods. Iowa State Univ. Press, Ames, Iowa. Received31 May 1990; accepted22 Oct. 1990.